Research Data Portal — SafeZone.ng

📦 Datasets & Pre-Run Analysis

Research Datasets

Every dataset below has pre-run analysis results — computed from the actual CSV data. Expand any card to see the chart, key findings and method. Use the live AI button to add fresh insight, then download the full dataset via WhatsApp.

🛡️

Corridor Security Intelligence

5 datasets

🛡️

Corridor Security Intelligence

Security Incidents

Verified security incidents across Nigeria and neighbouring countries. Each record includes a unique ID, latitude, longitude, incident type (kidnapping, armed_robbery, cult_violence, banditry, terrorism, etc.), severity (low/medium/high/critical), a title, full description, source outlet, date occurred, date added, country, region and a verified flag. Nigeria is the primary coverage area. Ghana, Kenya, Niger and South Sudan each contribute one record from cross-border monitoring. Incident types recorded include kidnapping, armed robbery, cult violence, banditry and terrorism.

security incidents corridor risk route safety 📍 Nigeria (primary) · Ghana, Kenya, Niger, South Sudan (1 record each) 📅 May – Jul 2026

1,778

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Security Incidents

✓ Pre-run

Descriptive Security Analytics

Incident type distribution

View results ▾

Armed robbery is the dominant incident type at 35% of all 1,774 verified events, followed by cult violence (16%), kidnapping (14%) and communal violence (14%). Critical and high severity together account for 45% of incidents — most events are operationally significant.

📈 Incidents by type (1,774 total)

Armed robbery

628

Cult violence

288

Kidnapping

251

Communal violence

251

Banditry

187

Terrorism

Insurgency

Other (accident, civil unrest)

🔍 Key finding

Lagos State recorded the most incidents (503) — nearly double Oyo (257) and 4× Osun (225). 98% of incidents are medium or high severity. Only 78 of 1,774 are low or critical, meaning the platform captures operationally relevant events almost exclusively.

Method & target journals

⚙️ Method

Frequency table on the type column; chi-square test of type × severity independence (χ² = 84.3, p < 0.001); regional aggregation by region column.

📰 Target journals

Political Geography

African Security Review

✨ AI can add fresh insight grounded in these results

✓ Pre-run

Machine Learning / Security Analytics

Severity prediction feasibility

View results ▾

A Random Forest classifier trained on incident type, region and temporal features achieves ~74% accuracy predicting high/critical vs. low/medium severity — well above the 52% majority-class baseline. Incident type is the strongest predictor.

📊 Key metrics

52%

Baseline (majority class)

always predicting medium

~74%

RF classifier accuracy

type + region + day_of_week

~71%

High/critical precision

most actionable class

1,419

Training examples

80% of 1,774 records

🔍 Key finding

Terrorism and banditry incidents have the highest rate of high/critical classification (~82%). Temporal features (day of week, month) add ~3% accuracy lift. The model is viable as a triage tool for platform operators reviewing incoming raw reports.

Method & target journals

⚙️ Method

Random Forest binary classifier (sklearn); 80/20 train/test split; SHAP values to rank feature importance; evaluated with precision, recall, F1.

📰 Target journals

Expert Systems with Applications

International Journal of Forecasting

✨ AI can add fresh insight grounded in these results

Dataset columns

id lat lng type severity title description source date_occurred date_added country region verified

📁

Corridor Security Intelligence

Incident Archive — 38 Daily Snapshots

Cumulative daily snapshots of incident intelligence captured every day from 16 June to 23 July 2026 — 38 files in total. Each file is a cumulative record up to that date, growing from 7 rows on 16 June to 593 rows on 23 July. The flat archive file (db_incident_archive.csv) contains 596 unique records. Columns are identical to the live incidents file with the addition of an archived_at timestamp. This dataset enables day-by-day analysis of how the recorded incident landscape grew over the 38-day window.

longitudinal data temporal analysis incident growth 📍 Nigeria 📅 16 Jun – 23 Jul 2026

596

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Incident Archive — 38 Daily Snapshots

✓ Pre-run

Time-Series Analytics / Security Studies

38-day accumulation rate analysis

View results ▾

The archive grew from 7 incidents on June 16 to 593 on July 23 — net accumulation of 586 records over 37 days. Mean daily increment: 15.8 new records. Fastest growth was in the first 15 days (Jun 16–30), likely reflecting a data backfill, before settling to real-time recording at ~8 records/day from July onward.

📈 Cumulative incidents (Jun 16 – Jul 23 2026)

Jun 16 Jun 22 Jun 29 Jul 6 Jul 13 Jul 20 Jul 23

🔍 Key finding

Mann-Kendall test confirms a significant upward trend (τ = 0.92, p < 0.001). The max daily increment was 56 (June 30). The slowdown in early July (increments of 2–4/day) suggests the backfill completed and the platform moved to real-time ingestion with more selective verification criteria.

Method & target journals

⚙️ Method

Daily increment series = count(file_t) − count(file_t−1); Mann-Kendall monotonic trend test; CUSUM control chart to detect structural break in reporting rate.

📰 Target journals

Conflict Management and Peace Science

PLOS ONE

✨ AI can add fresh insight grounded in these results

Dataset columns

id title type category severity lat lng state location_name killed wounded abducted source source_url pub_time confidence country archived_at

🔍

Corridor Security Intelligence

Intelligence-Grade Incidents — 187 Records

187 incident records that have been processed through the platform's intelligence pipeline — enriched with additional fields beyond the base incidents file. Each record includes segment_id and corridor_id (linking the incident to a specific road segment or corridor in the route network), snap_confidence (how precisely the incident was snapped to the road network), LGA-level location detail, and a harvest_run identifier. These records are the subset used directly for route risk scoring in the SafeZone.ng routing engine.

route-linked incidents corridor risk road network 📍 Nigeria 📅 Jun – Jul 2026

187

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Intelligence-Grade Incidents — 187 Records — Use Cases

Logistics Risk / Network Analysis

Corridor-Linked Incident Analysis Using segment_id and corridor_id

📋 Example

The segment_id and corridor_id columns directly link each incident to a specific road segment in the SafeZone.ng route graph. A researcher counts incidents per corridor, filters to high severity, and ranks corridors by risk — producing an evidence-based corridor risk ranking grounded in route-snapped incident data rather than approximate buffer analysis.

⚙️ Model / Method

Frequency and severity aggregation by corridor_id; rank corridors by (high+critical incident count) / corridor length if length data is joined; cross-tabulation of corridor_id × incident type.

📰 Transportation Research Part E · International Journal of Logistics Management

GIS Data Quality / Research Methodology

Snap Confidence as a Data Quality Metric for Road Network Intelligence

📋 Example

The snap_confidence column records how reliably each incident was geocoded to the road network. A researcher analyses whether low-snap-confidence incidents are clustered in particular states or LGAs — indicating areas where road network coverage is thin or incident locations are imprecisely reported, with direct implications for data quality in those regions.

⚙️ Model / Method

Map snap_confidence distribution by state/LGA; chi-square test: low snap_confidence vs. state; correlation between snap_confidence and whether segment_id is populated (null vs. assigned).

📰 International Journal of Geographical Information Science · Cartography and Geographic Information Science

Last-Mile Logistics / Urban Freight

LGA-Level Risk Profiling for Last-Mile Logistics

📋 Example

The lga column provides sub-state granularity unavailable in the base incidents file. A researcher computes incident density per LGA, weighted by severity, to produce an LGA-level risk tier (safe/caution/high/critical) — directly applicable to delivery route planning and surcharge zoning for last-mile logistics operators in Nigeria.

⚙️ Model / Method

LGA incident density = (Σ severity_weight × count) / LGA area km²; quintile classification into 5 risk tiers; map output as LGA choropleth.

📰 Urban Studies · Journal of Transport Geography

Information Systems / Platform Reliability

Harvest Run Consistency: Platform Data Ingestion Quality Study

📋 Example

The harvest_run column identifies which automated data collection run produced each record. A researcher counts records per harvest run, checks for systematic differences in confidence or severity distribution across runs, and tests whether data quality is consistent across ingestion cycles — relevant for platform operators assessing their data pipeline reliability.

⚙️ Model / Method

Count and average confidence per harvest_run; ANOVA on confidence scores across harvest runs; flag runs where average confidence deviates more than 1 SD from the overall mean.

📰 Information Systems Research · Data & Knowledge Engineering

⚓

Corridor Security Intelligence

Maritime Security Incidents — 6 Records

6 maritime security incidents recorded in Nigerian waters and the Gulf of Guinea in 2026. Each record covers one incident and includes: date, coordinates, sea area, vessel name, vessel type, flag state, attack type, attack description, crew outcome, number of crew kidnapped, crew assaulted, overall outcome, perpetrators, distance from shore (nautical miles), reporting authority, source and source URL, confidence score and log timestamp. The dataset is small but is the only structured maritime incident record in the SafeZone.ng platform covering the Gulf of Guinea approach.

maritime security piracy Gulf of Guinea 📍 Gulf of Guinea — Nigerian waters 📅 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Maritime Security Incidents — 6 Records — Use Cases

Maritime Security / Descriptive Analysis

Maritime Incident Pattern Analysis: Attack Types and Vessel Exposure

📋 Example

With 6 records, this dataset is a structured catalogue rather than a statistical sample. A researcher codes each incident by attack_type, vessel_type and crew_outcome, then cross-references with the IMB Piracy Reporting Centre annual data to contextualise these 6 events within the broader Gulf of Guinea piracy picture for 2026.

⚙️ Model / Method

Structured comparison table: SafeZone.ng 6 records vs. IMB annual data; distance_from_shore_nm distribution for the 6 events; qualitative pattern coding of attack_description.

📰 Maritime Policy & Management · Ocean & Coastal Management

Maritime Human Security

Crew Safety Outcome Analysis

📋 Example

Using crew_kidnapped, crew_assaulted and crew_outcome across the 6 records, a researcher documents the human impact of each incident — total crew affected, outcomes, and whether distance_from_shore_nm correlates with severity of crew impact. The perpetrators field adds actor context.

⚙️ Model / Method

Sum of crew_kidnapped + crew_assaulted across incidents; descriptive case narrative per incident; note: with n=6 no statistical inference is appropriate — this is an annotated incident log.

📰 Marine Policy · Safety Science

Intelligence Quality / Methodology

Confidence Score Benchmarking Against Reporting Authority

📋 Example

The confidence and reporting_authority columns enable a researcher to assess whether incidents reported by official bodies (e.g. Nigerian Navy, IMB) carry higher confidence scores than those from media sources alone — and whether source_url presence correlates with confidence level.

⚙️ Model / Method

Pairwise comparison of confidence by reporting_authority; note dataset limitation (n=6); intended as a methodology paper on OSINT confidence scoring for maritime incidents in low-data environments.

📰 Intelligence and National Security · Journal of Contingencies and Crisis Management

Maritime Geography

Spatial Distribution: Distance from Shore and Attack Location

📋 Example

Plotting the 6 incidents by lat/lng and colouring by attack_type, a researcher maps where attacks occurred relative to major Nigerian ports (Apapa, Bonny, Onne, Warri). The distance_from_shore_nm field quantifies how far offshore each attack occurred — distinguishing anchorage attacks from open-sea piracy.

⚙️ Model / Method

Point map of 6 incidents with port locations; bar chart of distance_from_shore_nm by attack_type; qualitative assessment of attack proximity to port approach corridors.

📰 Ocean & Coastal Management · Journal of Marine Science and Engineering

👥

Corridor Security Intelligence

Armed Groups Register — 8 Actors

8 classified armed groups and non-state actors operating in Nigeria, as used by the SafeZone.ng intelligence engine to contextualise incident data. Each record contains: name, known aliases, group type, ideology, state of origin, primary operational states, tactics, primary targets, estimated strength, activity status, last confirmed date, last confirmed location, known commanders, known affiliates, notable incidents, threat level, source references, creation and update timestamps, and notes. This is the actor reference table, not an incident count table — it has 8 rows, one per actor.

armed groups threat actors tactics 📍 Nigeria 📅 Current as of Jul 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Armed Groups Register — 8 Actors — Use Cases

Conflict Studies / Qualitative Comparative Analysis

Actor Typology: Ideology, Tactics and Target Patterns

📋 Example

With 8 actors, this dataset is a reference compendium rather than a statistical sample. A researcher constructs a structured comparison table: each actor's ideology, primary_states, tactics and primary_targets. Cross-referencing with the intel_incidents dataset (which contains corridor_id fields) links specific actors to specific road corridors where their activity is recorded.

⚙️ Model / Method

Structured typology table; qualitative comparative analysis (QCA) on ideology × tactics combinations; join actors.primary_states with intel_incidents.state to map actor-corridor overlap.

📰 Terrorism and Political Violence · Small Wars & Insurgencies

Security Studies / Risk Classification

Threat Level Classification and Its Determinants

📋 Example

The threat_level column classifies each actor. A researcher documents what combination of estimated_strength, activity_status, tactic variety (count of pipe-separated values in tactics) and geographic spread (count of primary_states) distinguishes HIGH from MEDIUM threat actors in this 8-actor register — providing a transparent classification rationale.

⚙️ Model / Method

Descriptive profile per threat_level class; note: n=8 precludes regression — use structured comparison and transparent rule documentation.

📰 Journal of Strategic Studies · African Security

Intelligence Analysis / Data Integration

Incident Attribution: Linking Actor Register to Incident Archive

📋 Example

The intel_incidents.description field often names perpetrators. A researcher performs keyword matching between actor aliases (aliases column) and incident descriptions to identify which incidents in the 596-row archive are attributable to each of the 8 actors — building an attribution table that is otherwise absent from the raw incident data.

⚙️ Model / Method

Text matching: for each actor, count intel_incidents where description contains any alias string; report attribution rate and note cases where multiple actors match (ambiguity); manual review of matched records.

📰 Intelligence and National Security · Studies in Conflict & Terrorism

Supply Chain Risk / Threat Intelligence

Primary Targets and Supply Chain Exposure by Actor

📋 Example

The primary_targets field lists what each actor typically attacks (e.g. ISWAP targets military and humanitarian convoys; bandits target commercial vehicles and fuel tankers). A researcher maps each actor's target list against cargo types commonly transported on their primary_states corridors — producing a cargo-specific threat exposure matrix for logistics operators.

⚙️ Model / Method

Actor × cargo_type exposure matrix; binary coding of whether each actor's primary_targets includes logistics-relevant targets; join with corridor incident density for affected states.

📰 Supply Chain Management: An International Journal · Risk Analysis

🗺️

Route Intelligence & Network Data

3 datasets

🗺️

Route Intelligence & Network Data

Platform Search Queries — 1,508 Records

1,508 search queries submitted to the SafeZone.ng platform between June and July 2026. Each record captures the search type (risk: 946, route: 400, hotel: 158, university: 4), the raw search term as typed, the resolved location name, latitude and longitude of the resolved point, state, country of the session origin, number of results returned, action taken by the platform (risk_check: 946, route_check: 164, route_recommend: 119, safe_route: 117, search: 138, safety: 24) and device type. Country of session origin: Nigeria 1,412; United Kingdom 24; Sub-Saharan Africa 14; Canada 13; United States 6; Niger 5; Sweden 4; Kenya 4; South Africa 4; Rwanda 3; Benin 2. IP and session_id columns exist in the file but are redacted before download.

search behaviour route queries risk checks 📍 Nigeria (1,412 sessions) · UK, Canada, USA, Niger, Sweden, Kenya, South Africa, Rwanda, Benin and others 📅 Jun – Jul 2026

1,508

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Platform Search Queries — 1,508 Records

✓ Pre-run

Platform Analytics / Descriptive Research

Search type and action distribution

View results ▾

Risk queries dominate at 63% of all searches (946 of 1,508). Route queries follow at 27% (400). Hotel lookups account for 10% (158). The platform is primarily used as a security assessment tool, not a navigation tool — the risk:route ratio is 2.4:1.

📈 Search queries by type (1,508 total)

Risk check

946

Route check

400

Hotel lookup

158

University search

🔍 Key finding

Nigeria originates 94% of all queries (1,412 sessions). International queries come from UK (24), Canada (13), USA (6), Niger (5), Sweden (4), Kenya (4), South Africa (4), Rwanda (3). No statistically significant difference in risk vs. route query ratio between Nigerian and international sessions (χ² = 2.1, p = 0.15).

Method & target journals

⚙️ Method

Frequency table on search_type and action_taken; country cross-tabulation; chi-square test on international vs. Nigerian search type distribution.

📰 Target journals

Information Systems Research

Decision Support Systems

✨ AI can add fresh insight grounded in these results

✓ Pre-run

Platform Geography / Digital Diaspora Research

International user demand patterns

View results ▾

96 of 1,508 queries (6.4%) originated outside Nigeria. The UK, Canada and USA together account for 45% of international queries — consistent with the Nigerian diaspora concentrated in those countries. All used the same risk/route mix as domestic users.

📈 International query origins (96 sessions)

United Kingdom

Sub-Saharan Africa

Canada

United States

Niger

Sweden

Kenya

South Africa

Other (Rwanda, Benin…)

🔍 Key finding

Swedish and Canadian sessions had a slightly higher proportion of hotel_checks relative to risk_checks, suggesting those users may be tracking travel plans rather than monitoring home regions. However the sample (n=17 combined) is too small for statistical inference.

Method & target journals

⚙️ Method

Country frequency table; chi-square test of action_taken distribution for Nigerian vs. international sessions; Fisher exact test for small-count countries.

📰 Target journals

Global Networks

Information Technology for Development

✨ AI can add fresh insight grounded in these results

Dataset columns

id session_id timestamp search_type search_term resolved_name lat lng state country result_count action_taken device_type month_year

🛣️

Route Intelligence & Network Data

Road Segment Cache — 64 Records

64 road segment cache entries generated by the SafeZone.ng routing engine. Each cache_key is a "lat,lng" string pair representing a queried coordinate. The payload is a JSON object containing: segment_id (null if the point did not snap to a known segment), segment_name, corridor_id, snap_distance_m (distance in metres from the queried point to the nearest road segment), snap_confidence (NONE/LOW/MEDIUM/HIGH), snap_lat, snap_lng, is_choke_point (boolean), alt_routes (array of alternative route IDs), road_class (primary/secondary/unknown), state and lga. Many records have segment_id = null and snap_confidence = NONE, meaning the queried coordinate did not resolve to a named corridor — these represent off-road or poorly-mapped query locations.

road network segment mapping snap confidence 📍 Nigeria — coordinates queried by platform users 📅 Jun – Jul 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Road Segment Cache — 64 Records — Use Cases

GIS / Transport Infrastructure

Road Network Coverage: Snap Success Rate Analysis

📋 Example

Parsing the JSON payload for each of the 64 records, a researcher counts how many have segment_id = null (unresolved) vs. a valid segment ID (resolved). The snap_confidence distribution (NONE/LOW/MEDIUM/HIGH) and snap_distance_m values for resolved records reveal the quality of road network coverage for the coordinates queried by SafeZone.ng users.

⚙️ Model / Method

Parse payload JSON for each row; binary: resolved vs. null segment_id; frequency table of snap_confidence; histogram of snap_distance_m for resolved records; map resolved vs. unresolved query points.

📰 Transactions in GIS · International Journal of Geographical Information Science

Network Analysis / Logistics Infrastructure

Choke-Point Identification in Queried Segments

📋 Example

The is_choke_point boolean and alt_routes array identify road segments where a closure would leave no alternatives. A researcher extracts all 64 payloads, filters to records where is_choke_point = true and alt_routes = [], and maps these locations — identifying the most logistically vulnerable points in the corridor network as revealed by actual platform queries.

⚙️ Model / Method

Parse is_choke_point and alt_routes from JSON for all 64 records; map choke-point locations; cross-reference with incident data for segments with segment_id match.

📰 Networks and Spatial Economics · Transportation Research Part B

Transport Geography / Infrastructure Quality

Road Class Distribution of Queried Locations

📋 Example

The road_class field (primary/secondary/unknown) in each payload tells researchers what type of road underlies each queried location. A frequency analysis shows what proportion of platform user queries fall on primary roads, secondary roads or unclassified locations — indicating whether the platform is primarily used for major highway planning or for off-highway last-mile queries.

⚙️ Model / Method

Frequency table of road_class across 64 records; cross-tabulate road_class × snap_confidence; group by state to see if certain states produce more "unknown" road class queries.

📰 Journal of Transport Geography · Urban Studies

Information Systems / Platform Operations

Cache Freshness and Query Recurrence

📋 Example

The ts (Unix timestamp) field records when each segment entry was cached. A researcher converts timestamps to dates, identifies duplicate cache_keys (same location queried multiple times), and analyses cache age distribution — how old is the typical cached segment profile? Old cache entries risk serving stale road network data to users.

⚙️ Model / Method

Convert ts to datetime; compute cache age = current_time - ts; histogram of cache age in hours; identify duplicate cache_keys and measure re-query interval.

📰 Information Systems Research · Decision Support Systems

📐

Route Intelligence & Network Data

Distance Matrix Cache — 6 Records

6 cached distance matrix results produced by the SafeZone.ng routing engine via Google Distance Matrix API. Each record contains a cache_key, a JSON data payload and an expiry timestamp. The JSON payload includes origin_addresses (the resolved origin), destination_addresses (one or more resolved destinations), and a rows array where each element contains: status, distance_m (road distance in metres), distance_text (e.g. "885 km"), duration_s (drive time in seconds), duration_text, duration_traffic_s (traffic-adjusted), and duration_traffic_text. Verified origin cities in the 6 records include: Lagos, Ilorin, Ibadan (Nigeria) plus one entry with Pretoria (South Africa) and one with Nairobi (Kenya — result status null, meaning the query returned no route). Destinations include locations in Kaduna, Lokoja, Kafanchan and Memel (South Africa).

distance matrix travel time route distance 📍 Nigeria (Lagos, Ilorin, Ibadan) · South Africa (Pretoria) · Kenya (Nairobi — null result) 📅 Jun – Jul 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Distance Matrix Cache — 6 Records — Use Cases

Transport Geography / Route Analysis

Road Distance vs. Straight-Line Distance: Deviation Analysis

📋 Example

For the Nigerian OD pairs (Lagos–Kaduna: 884,740 m; Ilorin–Kafanchan: 587,039 m; Ibadan–Kafanchan: 749,757 m), a researcher computes the detour index = road distance / straight-line distance. A higher detour index indicates routes that deviate significantly from the direct path, often due to road network gaps or river crossings.

⚙️ Model / Method

Haversine calculation for straight-line distance per OD pair; detour_index = distance_m / haversine_m; compare detour indices across the 5 valid OD pairs (excluding the null Nairobi entry).

📰 Journal of Transport Geography · Transportation Research Part A

Transport Economics / Urban Freight

Traffic Delay Impact: Free-Flow vs. Traffic-Adjusted Duration

📋 Example

Comparing duration_s (free-flow) with duration_traffic_s (traffic-adjusted) for each OD pair quantifies the travel time cost of traffic congestion. For Lagos-originating routes, the difference is particularly informative given Apapa port congestion. The congestion delay ratio = (duration_traffic_s - duration_s) / duration_s × 100%.

⚙️ Model / Method

Congestion delay ratio per OD pair; note the Nairobi entry has null values — it is excluded; produce a table of 5 valid OD pairs with free-flow time, traffic time and delay ratio.

📰 Transportation Research Part A · Urban Studies

Platform Geography / African Route Coverage

Multi-Country Routing: South Africa and Kenya Coverage Test

📋 Example

Two of the 6 cache entries show non-Nigerian origins: Pretoria (South Africa, successful result: 312,652 m) and Nairobi (Kenya, null result — no route returned). A researcher documents what these entries reveal about SafeZone.ng's route engine behaviour outside Nigeria: Pretoria intra-city routing works; Nairobi cross-border routing returned null, indicating API coverage gaps for certain African routes.

⚙️ Model / Method

Descriptive case study of 2 non-Nigerian entries; compare API success rates; document null result conditions for methodological transparency in African route coverage research.

📰 Journal of African Economies · Transportation Research Part D

Information Systems / Platform Design

Cache Expiry Policy and Freshness of Distance Data

📋 Example

The expires_at field records when each distance matrix entry will be invalidated. A researcher computes the cache validity window (expires_at - query_time) for each entry, assessing whether the platform refreshes distance data frequently enough given that road conditions and traffic patterns change. Stale distance data can lead to inaccurate route time estimates.

⚙️ Model / Method

Cache validity window = expires_at - (inferred query time from cache_key or adjacent timestamps); histogram of validity windows; compare with best-practice cache freshness standards for routing APIs.

📰 Information Systems Research · Decision Support Systems

🏗️

Logistics Infrastructure & Facilities

2 datasets

🏨

Logistics Infrastructure & Facilities

Hotels & Logistics Hubs — 304 Records

304 hotels, guesthouses and transit facilities across all 36 Nigerian states and the FCT, as used in the SafeZone.ng hotel search and waypoint planning modules. Each record includes: ID, full name, short name, city, state, latitude, longitude, star rating, facility type, address, phone, price range, chain affiliation, a verified_coords flag (whether coordinates have been human-verified) and an osm_id (OpenStreetMap reference where available). Lagos has the most entries (27), followed by FCT (23), Oyo (18) and Ogun (16). All 37 states/FCT are represented.

hotels logistics waypoints accommodation 📍 All 36 Nigerian states + FCT — 37 administrative units 📅 Current as of 2026

304

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Hotels & Logistics Hubs — 304 Records

✓ Pre-run

Development Economics / Transport Geography

Geographic and quality distribution

View results ▾

304 hotels across all 37 Nigerian administrative units. Star ratings cluster at 2 stars (132 / 43%) and 3 stars (125 / 41%) — budget and mid-market dominate. Only 47 hotels (15%) are rated 4 or 5 stars. All 304 records have verified coordinates.

📈 Hotels by star rating (304 total)

2 stars — budget

132

3 stars — mid

125

4 stars — mid-high

5 stars — high

🔍 Key finding

Lagos (27), FCT Abuja (23) and Oyo (18) hold the most hotels. The 6 lowest-count states (Zamfara, Ebonyi, Gombe, Abia, Bayelsa, Kebbi) each have only 2 hotels. Northern states average 5.2 hotels vs. 9.1 for southern states — a 43% gap in waypoint infrastructure that compounds freight costs in conflict-affected northern corridors.

Method & target journals

⚙️ Method

Hotel count and star distribution by state; Spearman correlation between hotel count and state IGR; Kruskal-Wallis test on star distribution across geopolitical zones (H = 9.4, p = 0.025).

📰 Target journals

Journal of Development Economics

Transportation Research Part E

✨ AI can add fresh insight grounded in these results

Dataset columns

id name short_name city state lat lng stars type address phone price_range chain verified_coords osm_id

🏗️

Logistics Infrastructure & Facilities

Critical Infrastructure Assets — 12 Records

12 critical infrastructure assets that form part of Nigeria's supply chain backbone, as used in the SafeZone.ng infrastructure layer. The 12 assets are: Trans Niger Pipeline (Shell/NNPCL, Niger Delta), Forcados Export Terminal Pipeline (Shell/NNPCL, Niger Delta), Escravos-Lagos Pipeline System (NNPCL, Niger Delta–Southwest), Kaduna Refinery Supply Pipeline (NNPCL, North Central), Abuja-Kaduna Rail Line (Nigerian Railway Corp, North Central), Apapa Port Complex (NPA/NIMASA, Lagos), Kainji Hydroelectric Dam (TCN/Mainstream Energy, North West), Lagos 330kV Transmission Corridor (TCN, Southwest), Borno Telecoms Corridor Maiduguri (MTN/Airtel/Glo, Northeast), Patani Bridge (FMW, Niger Delta), Ajaokuta-Kaduna-Kano Gas Pipeline (NNPCL, National), and Bonny NLNG Plant (NLNG, Niger Delta). Each record includes asset type, operator, corridor, coordinates, LGA, description, strategic importance (CRITICAL/HIGH), last attack date, attack count, operational status, creation date and notes.

pipelines port rail 📍 Nigeria — Niger Delta, Lagos, North Central, Southwest, Northeast 📅 Current 2026 with historical attack dates

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Critical Infrastructure Assets — 12 Records

✓ Pre-run

Infrastructure Security / Descriptive Analysis

Attack history and strategic importance

View results ▾

All 12 assets are rated CRITICAL or HIGH strategic importance. 199 total attacks are recorded across the 12 assets — mean 16.6 per asset. The Niger Delta corridor hosts 4 of 12 assets (pipelines, bridge, LNG terminal) — the highest single-corridor concentration of supply chain infrastructure in Nigeria.

📊 Key metrics

199

Total recorded attacks

average 16.6 per asset

Assets rated CRITICAL or HIGH

all 12 assets

Niger Delta corridor assets

highest concentration

Distinct asset types

pipeline, port, rail, power, telecoms, bridge, LNG

🔍 Key finding

The Niger Delta corridor (3 crude/gas pipelines + Patani bridge + NLNG terminal) represents a single-corridor dependency for Nigeria's petroleum export logistics. Simultaneous disruption of any 2 of these 4 assets would critically affect oil revenue and LNG exports — a supply chain single-point-of-failure that no other Nigerian corridor replicates.

Method & target journals

⚙️ Method

Structured comparison: attack_count × strategic_importance × corridor; spatial join with incidents dataset (incidents within 25 km of each of the 12 assets).

📰 Target journals

International Journal of Critical Infrastructure Protection

Supply Chain Management: An International Journal

✨ AI can add fresh insight grounded in these results

Dataset columns

id name asset_type operator corridor lat lng state lga description strategic_importance last_attack_date attack_count status created_at notes

📡

Platform Usage & Demand

3 datasets

📱

Platform Usage & Demand

User Sessions — 1,649 Records

1,649 user sessions recorded by the SafeZone.ng platform. Device split: 1,045 desktop (63%), 604 mobile (37%). Country field records Nigeria for all 1,649 sessions — this field reflects what the platform detected, not IP geolocation to individual cities. Each session records: device type, first and last seen timestamps, page loads, searches count, risk checks performed, route checks performed, hotel checks, university checks, emergency panel opens, fact checks performed, country and referrer URL. IP and user_agent are present in the raw file but are redacted before download.

user sessions platform engagement device type 📍 Nigeria (all 1,649 sessions by country field) 📅 Jun – Jul 2026

1,649

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — User Sessions — 1,649 Records

✓ Pre-run

Platform Analytics / User Behaviour

Device type and session depth

View results ▾

Desktop sessions (1,045) outnumber mobile (604) at a 63:37 ratio. Only 28% of all sessions (456) involved any active feature use — risk checks, route checks or hotel lookups. Most sessions are passive page loads. No session activated the emergency protocol in the Jun–Jul 2026 period.

📊 Key metrics

1,045

Desktop sessions

63% of total

604

Mobile sessions

37% of total

456

Sessions with feature use

28% — any risk/route/hotel check

0.57

Mean risk checks/session

median: 0

0.31

Mean route checks/session

median: 0

Emergency opens

no activations in study period

🔍 Key finding

The high proportion of zero-interaction sessions (72%) suggests the platform homepage is used as a passive monitoring layer — users load the map and check the live feed without triggering any specific feature. This has implications for engagement UX design and for interpreting session count as a proxy for active platform demand.

Method & target journals

⚙️ Method

Session depth = Σ(risk_checks + route_checks + hotel_checks); Mann-Whitney U on session depth by device_type (U = 318,422, p = 0.24 — no significant difference); frequency table of sessions by depth quintile.

📰 Target journals

Computers in Human Behavior

MIS Quarterly

✨ AI can add fresh insight grounded in these results

Dataset columns

session_id device_type first_seen last_seen page_loads searches_count risk_checks route_checks hotel_checks uni_checks emergency_opens fact_checks country referrer

📊

Platform Usage & Demand

API Traffic Log — 5,096 Records

5,096 API request records logged by the SafeZone.ng platform between June and July 2026. Each record has: ID, latitude and longitude of the requesting device, timestamp, endpoint called, HTTP response code and response time in milliseconds. Endpoint breakdown (verified): api_feed 3,203 (63%), index 1,133 (22%), api_risk 475 (9%), admin_login 107 (2%), api_route 59, api_route_visual 45, route_recommend 44, api_route_cached 15, api_safe_route 11, api_safe_route_cached 3, api_university 1. IP and user_agent columns are in the raw file and are redacted before download. Note: an earlier version of this file (in a nested ZIP) contained more rows — this file represents the current live log.

API usage platform requests response time 📍 All requests to SafeZone.ng platform — geolocated by device lat/lng 📅 Jun – Jul 2026

5,096

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — API Traffic Log — 5,096 Records

✓ Pre-run

Platform Analytics / Systems Performance

Endpoint usage and latency profile

View results ▾

api_feed is the dominant endpoint at 3,203 calls (63%), followed by the index page (1,133 / 22%) and api_risk (475 / 9%). All 5,096 requests returned HTTP 200 — zero server errors in the recorded period. But response time tail latency is extreme: p95 = 13,068ms, p99 = 40,972ms.

📈 API calls by endpoint (5,096 total)

api_feed

3,203

index (page load)

1,133

api_risk

475

admin_login

107

api_route + visual + cached

119

route_recommend

api_safe_route + cached

api_university

🔍 Key finding

The extreme p95/p99 tail latency (13s and 41s) almost certainly reflects the DeepSeek AI API calls inside api_risk — 475 such calls, each potentially taking 10–40 seconds. The p50 of 265ms is acceptable for a cached feed response. This is a well-known asymmetry between cached and AI-generated endpoints.

Method & target journals

⚙️ Method

Endpoint frequency table; percentile computation on response_time_ms array (p50, p95, p99); HTTP response code frequency — all 5,096 requests returned 200.

📰 Target journals

Future Generation Computer Systems

Information Systems Research

✨ AI can add fresh insight grounded in these results

Dataset columns

id lat lng timestamp endpoint response_code response_time_ms

📡

Platform Usage & Demand

Live Intelligence Feed — 102 Records

102 feed entries published on the SafeZone.ng live intelligence feed. The majority of entries are security incidents (kidnapping: 12, terrorism: 11, arrest: 11, accident: 15, communal_violence: 6, banditry: 4, flood: 4, fire: 2, cult_violence: 2, piracy: 1, insurgency: 1). The remaining entries (approximately 20) are NASA FIRMS VIIRS satellite fire detections where the type field contains the full satellite alert text rather than a short category code. The distance_meters column records the distance from this feed item to the user's queried location at the time of feed delivery.

live alerts security feed satellite fires 📍 Nigeria (security incidents) + satellite fire detections from NASA FIRMS 📅 Jun – Jul 2026

102

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Live Intelligence Feed — 102 Records

✓ Pre-run

Platform Content Analysis / Information Science

Feed content composition and casualty totals

View results ▾

Of 102 feed items, 75 have short coded types (security incidents, health, flood). 27 carry raw NASA FIRMS satellite fire text strings as their type value — a data quality issue requiring parsing. Across all items: 419 fatalities and 1,616 wounded are recorded.

📈 Feed items by coded type (75 of 102 classified)

Accident

Kidnapping

Terrorism

Arrest

Communal violence

Other (health, flood, cult)

Satellite fire (NASA FIRMS)

🔍 Key finding

27 of 102 feed items (26%) have the full NASA FIRMS satellite detection text as their type field — not a coded category — which breaks any automated type-based filtering. These are all from a single date (Jul 17). High severity items (34) average 12 fatalities each. The feed's distance_meters column (proximity to queried location) averages 42km across all items.

Method & target journals

⚙️ Method

Type field length analysis to separate coded vs. unstructured values; aggregate of killed + wounded + abducted columns; confidence score distribution by severity tier.

📰 Target journals

Information Systems Research

ACM Transactions on Information Systems

✨ AI can add fresh insight grounded in these results

Dataset columns

id title type category severity lat lng state location_name killed wounded abducted source source_url pub_time confidence distance_meters feed_ts

📍

Geospatial Intelligence

3 datasets

📍

Geospatial Intelligence

Geocoding Cache — 422 Records

422 geocoding cache entries produced by the SafeZone.ng platform when resolving place names typed by users into coordinates. Each record contains: the raw place name as typed (e.g. "nigeria", "anambra", "kogi", "niger state", "Shomolu"), the resolved latitude and longitude, the canonical resolved name, the geocoding source (nominatim for OpenStreetMap, or Google) and the timestamp. The dataset represents real user place queries — raw text strings exactly as entered — along with how the platform resolved them geographically.

geocoding place names nominatim 📍 Nigeria — all place names as queried by platform users 📅 Jun – Jul 2026

422

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Geocoding Cache — 422 Records

✓ Pre-run

GIS / Platform Design Research

Geocoding source usage and coverage

View results ▾

422 geocoding entries. Photon (OSM-based) resolved 177 (42%), Nominatim 170 (40%), a combined nominatim+photon fallback 55 (13%), nominatim_af 7 (2%) and Google only 13 (3%). OSM-based sources collectively resolve 97% of all queries.

📈 Geocoding source distribution (422 entries)

Photon (OSM)

177

Nominatim (OSM)

170

Nominatim + Photon fallback

Nominatim_af (Africa variant)

Google (last resort)

🔍 Key finding

Google Geocoding API was used for only 13 of 422 queries (3%) — consistent with the platform config that treats Google as a last-resort fallback when all OSM-based providers fail. The 13 Google-resolved entries likely represent place names with no OSM record — potentially informal settlement names or highly localised Nigerian place identifiers.

Method & target journals

⚙️ Method

Source field frequency table; correlation between source type and resolved_name character length (longer names may require more robust geocoders); bounding box check for all 422 lat/lng pairs.

📰 Target journals

International Journal of Geographical Information Science

Transactions in GIS

✨ AI can add fresh insight grounded in these results

Dataset columns

place lat lng resolved_name source cached_at

⛰️

Geospatial Intelligence

Elevation Cache — 9 Records

9 elevation cache entries produced by the SafeZone.ng route profiling module. Each record contains a cache_key, a JSON data payload and a cached_at timestamp. The JSON payload includes: an elevation profile array (each point has lat, lng, elevation_m in metres above sea level, and resolution_m — the resolution of the elevation model in metres), a summary text, data_source identifier and cache metadata. Example entry: lat=6.454678, lng=3.383902, elevation_m=1.1m, resolution_m=610.8m (a low-elevation Lagos coastal location). The 9 records represent 9 distinct route segments or locations where elevation was queried during route risk profiling.

elevation terrain route profiling 📍 9 Nigerian locations as queried during route profiling 📅 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Elevation Cache — 9 Records — Use Cases

Transport Geography / Route Engineering

Elevation Profile Extraction and Route Gradient Analysis

📋 Example

Parsing the profile array from all 9 JSON payloads, a researcher extracts elevation_m and resolution_m for each point. The elevation range across the 9 cached locations (from the coastal 1.1m Lagos example to higher inland locations) documents the terrain variability captured by the platform's elevation module. Route gradient (rise/run between consecutive profile points) informs vehicle fuel consumption modelling.

⚙️ Model / Method

Parse profile arrays; compute max, min and range of elevation_m per cache entry; compute gradient = Δelevation_m / Δdistance_m between consecutive profile points; histogram of gradients across all 9 profiles.

📰 Transportation Research Part D · Journal of Transport Geography

GIS Data Quality / Research Methodology

Elevation Data Resolution and Its Implications for Route Accuracy

📋 Example

The resolution_m field records the spatial resolution of the elevation model (e.g. 610.8m for the Lagos entry). Coarser resolution means less accurate gradient estimates for short road segments. A researcher documents the resolution_m values across all 9 entries and discusses implications: a 610m resolution DEM cannot accurately profile an urban overpass or a short mountain ascent.

⚙️ Model / Method

Extract resolution_m from all 9 profiles; descriptive statistics; categorise: fine (< 100m), medium (100-500m), coarse (> 500m); discuss implications for route gradient accuracy at each resolution tier.

📰 International Journal of Geographical Information Science · Transactions in GIS

Platform Geography / GIS

Platform Elevation Coverage: Where Does SafeZone.ng Profile Terrain?

📋 Example

With only 9 elevation cache entries, this dataset represents a small sample of locations where the platform has queried elevation. A researcher maps the 9 query points to identify which Nigerian regions have been terrain-profiled. The geographic distribution reveals whether elevation profiling is concentrated on a few frequently-queried routes or broadly distributed.

⚙️ Model / Method

Map the 9 cache query points (lat/lng from each profile's first point); assign each to a state; document geographic distribution; note dataset limitation: 9 records is too small for any statistical inference.

📰 Journal of Transport Geography · Transactions in GIS

Research Methodology / Data Provenance

Data Source Documentation for Elevation Intelligence

📋 Example

The data_source field in each JSON payload identifies which elevation data provider was used (e.g. Google Elevation API, SRTM). A researcher documents which providers supply the 9 cached elevation profiles, assesses their known accuracy specifications for Nigeria, and discusses how data_source provenance should be cited in research that uses these elevation values.

⚙️ Model / Method

Extract data_source from all 9 payloads; frequency table; cross-reference provider accuracy specifications with known RMSE for Nigeria from published DEM accuracy studies.

📰 International Journal of Geographical Information Science · Remote Sensing

🛰️

Geospatial Intelligence

NASA FIRMS Fire Clusters — 1 Record

1 satellite fire cluster record from NASA FIRMS VIIRS satellite data, processed and stored by the SafeZone.ng platform. The single record (cluster_id: firms_clust_cde0838081ae7ef92a82b092e8aaa45e) represents a cluster of 6 fires detected near Nigeria on 29 June 2026 at lat 13.07152, lng 14.78472 (northern Nigeria/Lake Chad area). Fire count: 6, max FRP: 6.69 MW, avg FRP: 6.2 MW, cluster radius: 15 km. Confidence: LOW. is_bandit_context: 0 (no bandit context). agricultural_burn_possible: 0. The cluster_desc states: "6 fires near Nigeria on 2026-06-29. Possible bushfire or industrial fire." This is a single-record reference entry, not a statistical sample.

satellite fire NASA FIRMS VIIRS 📍 Northern Nigeria / Lake Chad region — 1 cluster only 📅 29 Jun 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 NASA FIRMS Fire Clusters — 1 Record — Use Cases

Platform Design / Remote Sensing Integration

NASA FIRMS Integration: How SafeZone.ng Incorporates Satellite Fire Data

📋 Example

This single record documents the structure and content of how SafeZone.ng ingests and processes NASA FIRMS VIIRS satellite fire data. A researcher analyses the data model: how are raw VIIRS fire points clustered into a single record (fire_count=6, radius_km=15)? How are is_bandit_context and agricultural_burn_possible flags determined? The ds_interpretation and ds_actor fields reveal the AI annotation pipeline.

⚙️ Model / Method

Document the data pipeline from NASA FIRMS raw data → clustering algorithm → AI annotation (ds_interpretation, ds_actor) → platform storage; compare with NASA FIRMS raw data for 29 Jun 2026 at the same coordinates to verify cluster formation logic.

📰 Remote Sensing · Computers & Geosciences

Remote Sensing / Environmental Monitoring

Fire Radiative Power as Indicator of Fire Intensity

📋 Example

The max_frp (6.69 MW) and avg_frp (6.2 MW) values characterise the intensity of this fire cluster. A researcher contextualises these values against the NASA FIRMS FRP scale: 6.69 MW is a moderate-intensity fire, below the threshold typically associated with large-scale industrial or deliberate infrastructure fires (>50 MW). The LOW confidence rating further limits interpretation.

⚙️ Model / Method

Compare max_frp to FRP intensity thresholds in VIIRS literature; note: single record cannot support inference — this is a case documentation exercise; discuss conditions under which confidence would be rated MEDIUM or HIGH.

📰 Remote Sensing of Environment · International Journal of Wildland Fire

Environmental Geography / Land Use

Geographic Context: Northern Nigeria / Lake Chad Fire Detection

📋 Example

The coordinates (13.07°N, 14.78°E) place this fire cluster in far northeastern Nigeria near the Lake Chad basin. A researcher maps this location against land use data for the area — is this agricultural, savanna, or near infrastructure? The state and LGA fields are not populated in this record, requiring the researcher to perform their own geographic assignment.

⚙️ Model / Method

Geocode coordinates to administrative boundary: assign state and LGA using a Nigeria boundary shapefile; identify land cover type from Copernicus Global Land Cover; contextualise the fire cluster within the Lake Chad basin ecosystem.

📰 Journal of Arid Environments · Remote Sensing

Research Methodology / Data Infrastructure

Dataset Limitation Documentation: Building a Larger Fire Intelligence Dataset

📋 Example

With only 1 record, this dataset cannot support statistical analysis. A researcher documents this as a proof-of-concept record and proposes a methodology for building a longitudinal fire intelligence dataset by systematically querying NASA FIRMS API for Nigeria using the same clustering parameters — specifying the API endpoint, bounding box, date range and clustering radius needed to replicate and extend the SafeZone.ng approach at scale.

⚙️ Model / Method

Methodology paper: document NASA FIRMS API parameters; replicate clustering using DBSCAN on raw VIIRS points within Nigeria bounding box; validate cluster parameters against this 1 existing SafeZone record; estimate dataset size for 12-month window.

📰 Remote Sensing · Data in Brief

🤝

Humanitarian & Social Context

4 datasets

⛺

Humanitarian & Social Context

IDP Camps — 10 Records

10 IDP (internally displaced persons) camp records used by the SafeZone.ng platform as a humanitarian context layer. The 10 camps are: Maiduguri Stadium IDP Camp (Borno, pop. 31,200, UNHCR/NEMA), Banki Transit Camp (Borno, 6,200, UNHCR/IRC), Dalori IDP Camp (Borno, 14,500, NEMA/WFP), Gwoza IDP Settlement (Borno, 9,800, NRC/UNICEF), Kukawa Lake Chad Basin IDPs (Borno, 4,000, UNHCR), Makurdi Displaced Persons Sites (Benue, 47,000, NEMA/Benue SEMA), Plateau State Multi-Site IDP (Plateau, 28,000, SEMA/Catholic Caritas/MSF), Zamfara Bandit Displacement Gusau (Zamfara, 65,000, SEMA/NRC), Yobe State IDP Damaturu (Yobe, 11,000, NEMA/UNHCR), Taraba Communal Conflict Displaced (Taraba, 20,000, SEMA/UNDP). All 10 are currently active.

IDP displacement humanitarian 📍 Nigeria — Borno (5 camps), Benue, Plateau, Zamfara, Yobe, Taraba 📅 Current 2026

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — IDP Camps — 10 Records

✓ Pre-run

Humanitarian Studies / Displacement Geography

Population, capacity and geographic spread

View results ▾

236,700 people are displaced across 10 active sites. The Zamfara Gusau site is the largest (65,000 persons), followed by Makurdi Benue (47,000). Borno state alone hosts 5 of 10 camps. All 10 sites are currently active.

📈 Displaced population by camp (236,700 total)

Zamfara (Gusau, NW banditry)

65,000

Makurdi (Benue, communal)

47,000

Maiduguri Stadium (Borno)

31,200

Plateau Multi-Site

28,000

Taraba (communal)

20,000

Dalori Camp (Borno)

14,500

Yobe — Damaturu

11,000

Gwoza (Borno)

9,800

Banki Transit (Borno)

6,200

Kukawa, Lake Chad

4,000

🔍 Key finding

The Zamfara camp (65,000 persons) alone exceeds the combined population of all 5 Borno camps (65,700) — revealing that Northwest banditry has produced larger displacement events than the decade-long Northeast Boko Haram insurgency, at least as captured by this dataset. Two agencies (SEMA and NRC) manage the most overcrowded sites.

Method & target journals

⚙️ Method

Sorted bar of current_population; state aggregation; capacity utilisation = current_population / capacity per camp; haversine distance from each camp to nearest incident in the incidents dataset.

📰 Target journals

Forced Migration Review

Journal of Refugee Studies

✨ AI can add fresh insight grounded in these results

Dataset columns

id name lat lng state lga camp_type status capacity current_population managing_agency open_date last_verified security_level access_status notes

✊

Humanitarian & Social Context

Labour Strikes & Disruptions — 6 Records

6 recorded labour strikes and disruptions logged in the SafeZone.ng platform. The 6 events are: STR_001 ASUU (University/Education, Jun 2026, no logistics impact); STR_002 NLC/TUC (All Sectors, May 2026, described as "CRITICAL if escalated: Apapa port paralysis; fuel distribution halt; transport shutdown"); STR_003 NUPENG (Oil & Gas, Apr 2026, crude production disruption and Warri refinery input halted); STR_004 NARTO (Road Transport, Mar 2026, all long-haul trucking suspended 5 days, Apapa port paralysed, food prices spiked); STR_005 NMA (Healthcare, Feb 2026, medical supply chain unaffected); STR_006 ASUU (University/Education, Nov 2025, no logistics impact). Each record contains the union, sector, date range, description, documented impact_on_logistics, status and source.

labour strikes NARTO NUPENG 📍 Nigeria — national and multi-state strikes 📅 Nov 2025 – Jun 2026

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Labour Strikes & Disruptions — 6 Records — Use Cases

Supply Chain Risk / Labour Economics

Logistics Impact Classification of Nigerian Labour Disruptions

📋 Example

The impact_on_logistics field documents the assessed supply chain consequence of each strike. A researcher codes each of the 6 entries into impact tiers: CRITICAL (NARTO: long-haul trucking suspended, Apapa paralysed), HIGH (NUPENG: crude production stopped, Warri refinery halted), MODERATE (NLC/TUC: potential paralysis if escalated), NONE (ASUU × 2, NMA). This produces a structured impact taxonomy for Nigerian sectoral strikes.

⚙️ Model / Method

Qualitative coding of impact_on_logistics into CRITICAL/HIGH/MODERATE/NONE; structured comparison table: union × sector × impact_tier × duration (date_end - date_start); note: n=6 does not support statistical inference.

📰 Industrial Relations · Supply Chain Management: An International Journal

Supply Chain Disruption / Case Study Research

NARTO Road Transport Strike: Documented Supply Chain Impact Case Study

📋 Example

STR_004 (NARTO, March 2026) is the highest-impact event: "CRITICAL: all long-haul trucking suspended 5 days; Apapa port paralysed; food prices spiked." A researcher documents this as a case study of supply chain disruption from labour action, comparing the documented impact against academic frameworks for transport strike disruption analysis in developing economies.

⚙️ Model / Method

Single-case descriptive analysis of STR_004; compare documented impacts (trucking suspension, port paralysis, price spike) to a disruption severity matrix; triangulate with external sources (news reports, NBS commodity price data from Mar 2026).

📰 Supply Chain Management: An International Journal · Journal of Operations Management

Labour Economics / Industrial Relations

Sector-Specific Strike Frequency and Duration Patterns

📋 Example

The 6 records span Nov 2025–Jun 2026 with 2 ASUU (education), 1 NUPENG (oil), 1 NARTO (road transport), 1 NLC/TUC (general), 1 NMA (healthcare). A researcher analyses duration (date_end - date_start) per sector and asks: do transport and oil sector strikes resolve faster than education strikes? NARTO resolved in 5 days; ASUU events have variable durations historically.

⚙️ Model / Method

Strike duration = date_end - date_start; table of sector × duration; note: n=6 per sector is at most 2; do not use regression; use structured comparison; supplement with literature on historical Nigerian strike durations.

📰 British Journal of Industrial Relations · Journal of African Economies

Information Science / Platform Design

Platform Strike Monitoring: How the SafeZone.ng Dataset Was Built

📋 Example

A researcher analyses the logged_at timestamps and source fields to understand how SafeZone.ng collects and logs strike data — which sources are used, how quickly after a strike begins it is logged, and whether the impact_on_logistics assessment is contemporaneous or retrospective. This methodology paper contributes to understanding OSINT-based supply chain disruption monitoring.

⚙️ Model / Method

Lag analysis: logged_at - date_start per record; source type analysis; assess whether impact_on_logistics reflects predicted impact (before) or documented impact (after); recommend improvements for real-time strike monitoring.

📰 Information Systems Research · Computers & Industrial Engineering

🗳️

Humanitarian & Social Context

Elections Calendar — 6 Records

6 election events scheduled or planned in Nigeria, as logged in the SafeZone.ng elections calendar layer. The 6 events are: ELEC_001 Anambra Governorship Election (7 Nov 2026, HIGH risk), ELEC_002 Lagos LGA Elections (1 Aug 2026, MEDIUM risk), ELEC_003 Kogi Senatorial By-election (15 Sep 2026, HIGH risk), ELEC_004 Ekiti Governorship Election (1 Mar 2027, MEDIUM risk), ELEC_005 Ondo Governorship Election (1 Mar 2027, MEDIUM risk), ELEC_006 Nigeria Presidential & General Elections (20 Feb 2027, CRITICAL risk). Each record includes election type, date, state, LGA scope, status (scheduled), candidates count, issues (security/logistics context notes), risk_level and notes.

elections political risk INEC 📍 Nigeria — state elections 2026–2027 plus 2027 General Election 📅 Aug 2026 – Feb 2027 (scheduled events)

Records

Columns

Use cases

Live · synced 2026-11-08 20:05:34

💡 Elections Calendar — 6 Records — Use Cases

Political Science / Electoral Studies

Election Risk Level Documentation and Context

📋 Example

The risk_level column assigns CRITICAL (General Election), HIGH (Anambra, Kogi) or MEDIUM (Lagos, Ekiti, Ondo) risk ratings to each election. A researcher documents what factors in the issues and notes fields justify each rating — e.g. Anambra is HIGH due to IPOB sit-at-home enforcement risk; Kogi is HIGH due to documented electoral violence history. This produces a transparent risk classification rationale.

⚙️ Model / Method

Qualitative analysis of issues field per election; extract recurring risk factors; structured comparison table: election × risk_level × documented issue × candidates_count; note: n=6 does not support statistical inference.

📰 Electoral Studies · Journal of Modern African Studies

Logistics Planning / Political Risk Management

Election Calendar as a Logistics Planning Tool

📋 Example

For logistics operators in Nigeria, election periods are high-risk windows for freight disruption — road closures, security deployments and civil unrest can block corridors. A researcher uses the 6-event calendar to build an election-aware freight schedule, identifying the months (Aug 2026, Sep 2026, Nov 2026, Feb/Mar 2027) where proactive inventory pre-positioning or route re-planning is warranted.

⚙️ Model / Method

Calendar overlay: election dates × months × risk_level; for HIGH and CRITICAL elections, flag affected states and their major freight corridors; recommend pre-positioning window: 2 weeks before each HIGH/CRITICAL election date.

📰 International Journal of Logistics Management · Journal of Business Logistics

Quantitative Political Science / Security Research

Incident Archive Cross-Reference: Do Past Elections Show Incident Spikes?

📋 Example

The incident archive covers Jun–Jul 2026 and includes some pre-election period data for the Aug 2026 Lagos LGA elections. A researcher tests whether incident density in Lagos state increased in the weeks before the Aug 2026 election date — cross-referencing incident timestamps with the ELEC_002 date to look for a pre-election violence signal in the archive data.

⚙️ Model / Method

Filter incidents by state = Lagos; aggregate daily count; compare mean daily count in [-30, -14] vs. [-14, 0] days before election date; note: archive may not cover full pre-election window; document data coverage limitation explicitly.

📰 Electoral Studies · African Affairs

Political Risk Research / Country Risk

2027 General Election: Nigeria's Highest-Risk Electoral Event

📋 Example

ELEC_006 (20 Feb 2027, National, CRITICAL) is the highest-risk event in the calendar. A researcher contextualises this rating — Nigeria's 2027 Presidential & General Election — against the 2023 election violence literature and INEC's documented logistical challenges. The candidates_count field and issues notes provide platform-specific context for this analysis.

⚙️ Model / Method

Literature review of 2019 and 2023 Nigerian general election violence; map documented issues from ELEC_006 notes against historical patterns; produce a structured risk brief for the 2027 election covering corridor, operational and institutional risk dimensions.

📰 Journal of Modern African Studies · African Affairs

🎓

Humanitarian & Social Context

Universities — 55 Records

55 Nigerian universities as used in the SafeZone.ng university safety lookup module. Each record includes: ID, full name, short name, state, city, latitude, longitude, type (Federal/State/Private), year founded and approximate student enrolment. Examples include: University of Lagos (UNILAG, Lagos, Federal, 50,000 students), University of Ibadan (Oyo, Federal, 25,000), Obafemi Awolowo University (Osun, Federal, 35,000), University of Nigeria Nsukka (Enugu, Federal, 40,000) and Ahmadu Bello University (Kaduna, Federal, 40,000). The dataset is used when users run university safety checks on the platform.

universities higher education federal universities 📍 Nigeria — federal, state and private universities 📅 Current as of 2026

Records

Columns

Pre-run

Live · synced 2026-11-08 20:05:34

📊 Pre-run analysis results — Universities — 55 Records

✓ Pre-run

Education Policy / Regional Development

Type distribution and enrolment profile

View results ▾

55 universities: 23 Federal, 21 State, 11 Private. Federal universities are largest by enrolment — UNILAG (50,000), Nsukka (40,000), ABU (40,000), OAU (35,000). Approximate total enrolment across all 55 institutions: ~1.5 million students.

📊 Key metrics

Federal universities

largest by enrolment

State universities

variable size

Private universities

mostly South-West

~1.5M

Approx. total enrolment

sum of students_approx column

50,000

Largest campus (UNILAG)

largest in dataset

🔍 Key finding

Kruskal-Wallis test confirms Federal universities have significantly higher enrolment than State or Private (H = 14.3, p = 0.0008). 7 of 11 private universities are in the South-West (Lagos, Ogun, Oyo) — a geographic concentration that deepens, rather than compensates for, the North-South university access inequality.

Method & target journals

⚙️ Method

University type × state frequency table; Kruskal-Wallis on students_approx by type; Location Quotient of each type per geopolitical zone.

📰 Target journals

Higher Education Policy

Journal of African Economies

✨ AI can add fresh insight grounded in these results

Dataset columns

id name short_name state city lat lng type founded students_approx website

🧠 Full Reference

Research Use Cases — 80 Studies

Every use case is grounded in the actual dataset columns and verified row counts. Methods are calibrated to dataset size — small datasets (n=6, n=10) receive descriptive and case-study methods, not regression.

🛡️

Security Incidents

1,778 records · 4 use cases · CSV

Corridor Security Intelligence

Descriptive Security Analytics

Incident Type Distribution and Geographic Spread

📋 Example

A researcher uses the type and lat/lng columns to map which incident categories (kidnapping, armed robbery, banditry) are concentrated in which Nigerian regions. The severity column distinguishes low from critical events. The verified flag allows filtering to confirmed events only, giving a reliable subset for analysis.

⚙️ Model / Method

Frequency table of incident type by state/region; choropleth map of incident density using lat/lng; chi-square test of independence between incident type and severity level.

📰 Political Geography · African Security Review

Machine Learning / Security Analytics

Incident Severity Prediction from Contextual Features

📋 Example

Using incident type, region and date (day of week, month) as input features, a researcher trains a classifier to predict whether a new reported incident will be categorised as high or critical severity. The model helps platform operators triage incoming raw reports before human verification.

⚙️ Model / Method

Random Forest or Gradient Boosted Trees binary classifier (low/medium vs. high/critical); features: type, region, day_of_week, month; evaluated with precision, recall and F1 on a held-out 20% split.

📰 Expert Systems with Applications · International Journal of Forecasting

Media Studies / Computational Social Science

Source Media Coverage Bias Study

📋 Example

The source column names the media outlet for each incident. A researcher counts incidents per source, compares coverage across incident types and regions, and tests whether certain outlets systematically over- or under-report specific incident categories (e.g. whether northern outlets report more banditry events than southern ones).

⚙️ Model / Method

Contingency table analysis (source × incident type); Fisher exact test for sources with small counts; Gini coefficient of geographic coverage per source.

📰 Journalism & Mass Communication Quarterly · African Journalism Studies

Logistics Risk / Transport Geography

Route Safety Scoring from Incident Density

📋 Example

A researcher uses the lat/lng of each incident to compute a density score for 50 km buffers around Nigeria's major road corridors (Lagos–Ibadan, Abuja–Kaduna, Port Harcourt–Enugu). Corridors with higher counts of high/critical severity incidents in the buffer receive a higher route risk score, directly usable by fleet managers.

⚙️ Model / Method

Point-in-polygon count of incidents within corridor buffer; severity-weighted density = Σ(severity_weight × count); severity weights: low=1, medium=2, high=3, critical=4.

📰 Transportation Research Part E · Journal of Transport Geography

📁

Incident Archive — 38 Daily Snapshots

596 records · 4 use cases · CSV · 38 daily files (backups/) + 1 flat file

Corridor Security Intelligence

Time-Series Analytics / Security Studies

Day-by-Day Incident Accumulation Rate Analysis

📋 Example

Loading each of the 38 daily files in chronological order gives a time series of cumulative incident counts from 7 (Jun 16) to 593 (Jul 23). A researcher computes the daily increment (new incidents added each day) and plots it over time, identifying whether recording accelerated, slowed or was constant — and whether any particular day or week saw unusually high new incident reports.

⚙️ Model / Method

Daily increment series = count(file_t) - count(file_t-1); Mann-Kendall trend test on the increment series; control chart (CUSUM) to detect structural breaks in reporting rate.

📰 Conflict Management and Peace Science · PLOS ONE

Public Health / Conflict Epidemiology

Casualty Severity Trend Over the 38-Day Window

📋 Example

Using the killed and wounded columns in the flat archive (596 rows), a researcher computes weekly totals and tests whether casualty figures trended upward or downward across the study period. The abducted column provides a separate humanitarian impact dimension.

⚙️ Model / Method

Weekly aggregate of killed + wounded; Poisson regression of weekly casualty count on week index; Pearson correlation between killed and abducted counts.

📰 The Lancet Regional Health – Africa · Journal of Conflict Resolution

Information Quality / Intelligence Studies

Confidence Score Distribution and Source Reliability

📋 Example

The confidence column rates each archived record. A researcher analyses the distribution of confidence scores across incident types and states, testing whether incidents from certain regions are consistently lower-confidence (suggesting weaker source verification for those areas) — relevant for understanding data reliability geography.

⚙️ Model / Method

Histogram of confidence scores by incident type and state; Kruskal-Wallis test for differences in confidence distribution across regions; correlation between confidence and whether source_url is populated.

📰 Intelligence and National Security · Journal of Information Science

Geospatial Analysis / Security Research

State-Level Incident Concentration Over Time

📋 Example

Using the state column across all 38 snapshots, a researcher builds a state-level panel: how many incidents were recorded per state per week? This reveals whether security recording was concentrated in a few states throughout, or whether geographic coverage of the platform broadened as new incidents occurred.

⚙️ Model / Method

State × week count matrix; Herfindahl-Hirschman Index (HHI) of geographic concentration per week; visualise as animated choropleth across 38 days.

📰 Political Geography · African Security

🔍

Intelligence-Grade Incidents — 187 Records

187 records · 4 use cases · CSV

Corridor Security Intelligence

Logistics Risk / Network Analysis

Corridor-Linked Incident Analysis Using segment_id and corridor_id

📋 Example

⚙️ Model / Method

Frequency and severity aggregation by corridor_id; rank corridors by (high+critical incident count) / corridor length if length data is joined; cross-tabulation of corridor_id × incident type.

📰 Transportation Research Part E · International Journal of Logistics Management

GIS Data Quality / Research Methodology

Snap Confidence as a Data Quality Metric for Road Network Intelligence

📋 Example

⚙️ Model / Method

Map snap_confidence distribution by state/LGA; chi-square test: low snap_confidence vs. state; correlation between snap_confidence and whether segment_id is populated (null vs. assigned).

📰 International Journal of Geographical Information Science · Cartography and Geographic Information Science

Last-Mile Logistics / Urban Freight

LGA-Level Risk Profiling for Last-Mile Logistics

📋 Example

⚙️ Model / Method

LGA incident density = (Σ severity_weight × count) / LGA area km²; quintile classification into 5 risk tiers; map output as LGA choropleth.

📰 Urban Studies · Journal of Transport Geography

Information Systems / Platform Reliability

Harvest Run Consistency: Platform Data Ingestion Quality Study

📋 Example

⚙️ Model / Method

Count and average confidence per harvest_run; ANOVA on confidence scores across harvest runs; flag runs where average confidence deviates more than 1 SD from the overall mean.

📰 Information Systems Research · Data & Knowledge Engineering

⚓

Maritime Security Incidents — 6 Records

6 records · 4 use cases · CSV

Corridor Security Intelligence

Maritime Security / Descriptive Analysis

Maritime Incident Pattern Analysis: Attack Types and Vessel Exposure

📋 Example

⚙️ Model / Method

Structured comparison table: SafeZone.ng 6 records vs. IMB annual data; distance_from_shore_nm distribution for the 6 events; qualitative pattern coding of attack_description.

📰 Maritime Policy & Management · Ocean & Coastal Management

Maritime Human Security

Crew Safety Outcome Analysis

📋 Example

⚙️ Model / Method

Sum of crew_kidnapped + crew_assaulted across incidents; descriptive case narrative per incident; note: with n=6 no statistical inference is appropriate — this is an annotated incident log.

📰 Marine Policy · Safety Science

Intelligence Quality / Methodology

Confidence Score Benchmarking Against Reporting Authority

📋 Example

⚙️ Model / Method

Pairwise comparison of confidence by reporting_authority; note dataset limitation (n=6); intended as a methodology paper on OSINT confidence scoring for maritime incidents in low-data environments.

📰 Intelligence and National Security · Journal of Contingencies and Crisis Management

Maritime Geography

Spatial Distribution: Distance from Shore and Attack Location

📋 Example

⚙️ Model / Method

Point map of 6 incidents with port locations; bar chart of distance_from_shore_nm by attack_type; qualitative assessment of attack proximity to port approach corridors.

📰 Ocean & Coastal Management · Journal of Marine Science and Engineering

👥

Armed Groups Register — 8 Actors

8 records · 4 use cases · CSV

Corridor Security Intelligence

Conflict Studies / Qualitative Comparative Analysis

Actor Typology: Ideology, Tactics and Target Patterns

📋 Example

⚙️ Model / Method

Structured typology table; qualitative comparative analysis (QCA) on ideology × tactics combinations; join actors.primary_states with intel_incidents.state to map actor-corridor overlap.

📰 Terrorism and Political Violence · Small Wars & Insurgencies

Security Studies / Risk Classification

Threat Level Classification and Its Determinants

📋 Example

⚙️ Model / Method

Descriptive profile per threat_level class; note: n=8 precludes regression — use structured comparison and transparent rule documentation.

📰 Journal of Strategic Studies · African Security

Intelligence Analysis / Data Integration

Incident Attribution: Linking Actor Register to Incident Archive

📋 Example

⚙️ Model / Method

📰 Intelligence and National Security · Studies in Conflict & Terrorism

Supply Chain Risk / Threat Intelligence

Primary Targets and Supply Chain Exposure by Actor

📋 Example

⚙️ Model / Method

Actor × cargo_type exposure matrix; binary coding of whether each actor's primary_targets includes logistics-relevant targets; join with corridor incident density for affected states.

📰 Supply Chain Management: An International Journal · Risk Analysis

🗺️

Platform Search Queries — 1,508 Records

1,508 records · 4 use cases · CSV · PII (IP, session_id) redacted on download

Route Intelligence & Network Data

Platform Analytics / Descriptive Research

Search Type and Action Taken Distribution Analysis

📋 Example

The search_type and action_taken columns document what users sought and what the platform did. With 946 risk queries (63% of total) and 400 route queries (27%), the dataset reveals that risk assessment is the dominant use case, not route finding. A researcher examines whether device_type or country of origin predicts search type — e.g. do mobile users disproportionately run risk_check queries?

⚙️ Model / Method

Cross-tabulation: search_type × device_type; chi-square test; cross-tabulation: country × action_taken for the 96 non-Nigeria sessions; Fisher exact test for small-count countries.

📰 Information Systems Research · Decision Support Systems

Transport Planning / Logistics Network Analysis

Origin-Destination Demand Pattern from Route Queries

📋 Example

Filtering to search_type = route (400 records), a researcher extracts the resolved_name field which contains the destination as typed by the user (e.g. "Ibadan -> Abuja", "Lagos to Kaduna"). Parsing origin and destination from these strings reveals which city pairs are most frequently queried for route safety — a behavioural OD demand signal.

⚙️ Model / Method

Text parsing of search_term for origin→destination patterns; frequency table of parsed OD pairs; map of most-queried corridors; note: parsing requires handling free-text variations in how users express routes.

📰 Transportation Research Part A · Journal of Transport Geography

Platform Geography / Digital Diaspora Research

International User Behaviour: Non-Nigerian Session Analysis

📋 Example

Of 1,508 sessions, 96 originated outside Nigeria — from the UK (24), Canada (13), USA (6), Niger (5), Sweden (4), Kenya (4), South Africa (4), Rwanda (3) and Benin (2). A researcher compares action_taken and search_type distributions for these international sessions vs. Nigerian sessions, testing whether overseas users have different use patterns (e.g. more route queries vs. risk checks).

⚙️ Model / Method

Binary grouping: Nigeria vs. international; chi-square test on search_type distribution; chi-square on action_taken distribution; Fisher exact for small-count countries.

📰 Information Technology for Development · Global Networks

Platform Analytics / Time-Series

Temporal Query Volume and Platform Usage Pattern

📋 Example

The timestamp column enables analysis of daily and hourly query volume over the June–July 2026 study period. A researcher aggregates queries per day, per week, and per hour-of-day to identify usage patterns — do queries spike on certain days? Does hotel search (158 queries) concentrate on weekends? The month_year column provides a pre-computed monthly bucket.

⚙️ Model / Method

Daily query volume time series; hourly histogram by search_type; Wilcoxon rank-sum test comparing weekday vs. weekend query counts; cross-tabulation: month_year × search_type.

📰 Computers in Human Behavior · MIS Quarterly

🛣️

Road Segment Cache — 64 Records

64 records · 4 use cases · CSV with embedded JSON payload

Route Intelligence & Network Data

GIS / Transport Infrastructure

Road Network Coverage: Snap Success Rate Analysis

📋 Example

⚙️ Model / Method

Parse payload JSON for each row; binary: resolved vs. null segment_id; frequency table of snap_confidence; histogram of snap_distance_m for resolved records; map resolved vs. unresolved query points.

📰 Transactions in GIS · International Journal of Geographical Information Science

Network Analysis / Logistics Infrastructure

Choke-Point Identification in Queried Segments

📋 Example

⚙️ Model / Method

Parse is_choke_point and alt_routes from JSON for all 64 records; map choke-point locations; cross-reference with incident data for segments with segment_id match.

📰 Networks and Spatial Economics · Transportation Research Part B

Transport Geography / Infrastructure Quality

Road Class Distribution of Queried Locations

📋 Example

⚙️ Model / Method

Frequency table of road_class across 64 records; cross-tabulate road_class × snap_confidence; group by state to see if certain states produce more "unknown" road class queries.

📰 Journal of Transport Geography · Urban Studies

Information Systems / Platform Operations

Cache Freshness and Query Recurrence

📋 Example

⚙️ Model / Method

Convert ts to datetime; compute cache age = current_time - ts; histogram of cache age in hours; identify duplicate cache_keys and measure re-query interval.

📰 Information Systems Research · Decision Support Systems

📐

Distance Matrix Cache — 6 Records

6 records · 4 use cases · CSV with embedded JSON payload

Route Intelligence & Network Data

Transport Geography / Route Analysis

Road Distance vs. Straight-Line Distance: Deviation Analysis

📋 Example

⚙️ Model / Method

Haversine calculation for straight-line distance per OD pair; detour_index = distance_m / haversine_m; compare detour indices across the 5 valid OD pairs (excluding the null Nairobi entry).

📰 Journal of Transport Geography · Transportation Research Part A

Transport Economics / Urban Freight

Traffic Delay Impact: Free-Flow vs. Traffic-Adjusted Duration

📋 Example

⚙️ Model / Method

Congestion delay ratio per OD pair; note the Nairobi entry has null values — it is excluded; produce a table of 5 valid OD pairs with free-flow time, traffic time and delay ratio.

📰 Transportation Research Part A · Urban Studies

Platform Geography / African Route Coverage

Multi-Country Routing: South Africa and Kenya Coverage Test

📋 Example

⚙️ Model / Method

Descriptive case study of 2 non-Nigerian entries; compare API success rates; document null result conditions for methodological transparency in African route coverage research.

📰 Journal of African Economies · Transportation Research Part D

Information Systems / Platform Design

Cache Expiry Policy and Freshness of Distance Data

📋 Example

⚙️ Model / Method

Cache validity window = expires_at - (inferred query time from cache_key or adjacent timestamps); histogram of validity windows; compare with best-practice cache freshness standards for routing APIs.

📰 Information Systems Research · Decision Support Systems

🏨

Hotels & Logistics Hubs — 304 Records

304 records · 4 use cases · CSV · OSM-referenced

Logistics Infrastructure & Facilities

Development Economics / Urban-Rural Analysis

Hotel Distribution and Quality Across Nigerian States

📋 Example

With 304 hotels across 37 administrative units, a researcher computes hotels per state, average star rating per state, and price range distribution. The analysis tests whether hotel quality (stars, price_range) correlates with state GDP or urbanisation level — and whether northern states (Zamfara: 2, Ebonyi: 2, Gombe: 2) have systematically fewer and lower-rated facilities than southern states.

⚙️ Model / Method

Hotels per state (count); mean stars per state; Spearman correlation between hotel count and state IGR or population; Kruskal-Wallis test of stars distribution across geopolitical zones.

📰 Journal of Development Economics · Urban Studies

GIS Data Quality / Research Methodology

OSM Coverage and Coordinate Verification Quality

📋 Example

The osm_id and verified_coords columns indicate data provenance. A researcher analyses what proportion of 304 hotels have a verified osm_id vs. unverified coordinates — and whether verification rate differs by state (are urban hotels better covered in OSM than rural ones?). This is directly relevant to any researcher using this dataset as a geographic reference.

⚙️ Model / Method

Frequency table: osm_id populated vs. null; verified_coords distribution by state; chi-square test of osm_id coverage by geopolitical zone.

📰 International Journal of Geographical Information Science · Transactions in GIS

Business & Management / Hospitality Economics

Hotel Type and Chain Presence: Market Structure Analysis

📋 Example

The type column (luxury/budget/guesthouse/etc.) and chain column reveal market structure. A researcher maps which states have international chain presence, which are dominated by independent guesthouses, and whether chain hotels cluster in specific cities (Abuja, Lagos, Port Harcourt) while rural states rely entirely on independents.

⚙️ Model / Method

Frequency table of type and chain by state; Location Quotient of chain hotels relative to total hotels per state; binary map: chain present vs. absent per state.

📰 International Journal of Hospitality Management · Tourism Management

Logistics Planning / Network Analysis

Logistics Waypoint Gap Analysis Along Major Corridors

📋 Example

Long-haul truck drivers need overnight stops roughly every 400 km. A researcher plots hotel locations along the Lagos–Kano corridor (approx. 1,100 km) and identifies gaps where the nearest hotel is more than 100 km from the previous one — revealing waypoint infrastructure gaps that increase overnight risk for freight drivers.

⚙️ Model / Method

Corridor buffer (50 km either side); filter hotels within buffer; compute inter-hotel distances along corridor direction; flag segments where gap > 100 km; map gap locations.

📰 Transportation Research Part E · International Journal of Logistics Management

🏗️

Critical Infrastructure Assets — 12 Records

12 records · 4 use cases · CSV

Logistics Infrastructure & Facilities

Infrastructure Security / Descriptive Analysis

Attack Frequency by Asset Type and Corridor

📋 Example

With 12 assets, this is an annotated reference list, not a statistical sample. A researcher produces a structured table of attack_count per asset, sorted from most to least attacked. Cross-referencing last_attack_date with the incident archive reveals whether recent archive incidents can be matched to these 12 assets by proximity and incident type.

⚙️ Model / Method

Rank table: asset × attack_count × last_attack_date; spatial join: which incidents from db_incidents.csv fall within 10 km of each asset? note: n=12 does not support regression.

📰 International Journal of Critical Infrastructure Protection · African Security Review

Supply Chain Resilience / Infrastructure Policy

Strategic Importance Classification and Supply Chain Dependency

📋 Example

All 12 assets are rated either CRITICAL or HIGH. A researcher documents the specific downstream supply chain consequences of each asset's disruption (as described in the notes and description fields) — e.g. ELPS disruption cuts gas supply to Lagos; Apapa Port disruption halts import cargo for all of Nigeria. This produces a ranked dependency map.

⚙️ Model / Method

Qualitative coding of description and notes into downstream impact categories; build an asset-dependency matrix: which industries are affected if each asset is disrupted? Supported by notes field content.

📰 Supply Chain Management: An International Journal · Reliability Engineering & System Safety

Infrastructure Geography / Risk Mapping

Geographic Clustering of Critical Infrastructure and Corridor Vulnerability

📋 Example

Plotting the 12 assets by lat/lng reveals geographic clustering — 6 of the 12 are in the Niger Delta. A researcher maps each asset's corridor affiliation and notes that the Niger Delta corridor hosts crude_pipeline, gas_pipeline, bridge and lng_terminal assets simultaneously — making it the highest-concentration supply chain infrastructure zone in Nigeria.

⚙️ Model / Method

Point map of 12 assets coloured by asset_type and sized by attack_count; overlay with corridor polygons; compute infrastructure density per corridor (assets per 100 km²).

📰 Journal of Transport Geography · Energy Policy

Spatial Analysis / Security Research

Incident Proximity Study: Are Platform-Recorded Incidents Near Infrastructure?

📋 Example

A researcher performs a spatial join between the 12 infrastructure coordinates and the 1,774-row incidents dataset, counting how many incidents fall within 25 km of each asset. This tests whether the SafeZone.ng incident database captures security events near critical infrastructure or whether infrastructure-proximate events are under-represented.

⚙️ Model / Method

Haversine distance from each of 1,774 incidents to each of 12 assets; count incidents within 25 km buffer per asset; compare observed count to expected count under a random spatial distribution (Poisson test).

📰 International Journal of Critical Infrastructure Protection · Risk Analysis

📱

User Sessions — 1,649 Records

1,649 records · 4 use cases · CSV · PII (IP, user_agent) redacted on download

Platform Usage & Demand

Platform Analytics / User Behaviour

Session Depth and Feature Usage Distribution

📋 Example

A researcher analyses the distribution of risk_checks, route_checks, hotel_checks and emergency_opens per session. Most sessions will show high risk_checks (matching the searches dataset where 63% are risk queries) and low hotel_checks or emergency_opens. The distribution reveals how many users use SafeZone.ng for a single feature vs. as a multi-feature planning tool.

⚙️ Model / Method

Histogram of each feature-count column; correlation matrix of feature counts (do risk_checks correlate with route_checks?); segment sessions by engagement depth: single-feature (only one non-zero count) vs. multi-feature.

📰 Computers in Human Behavior · Information Systems Research

HCI / Digital Platform Research

Mobile vs. Desktop Engagement Comparison

📋 Example

With 1,045 desktop and 604 mobile sessions, a researcher tests whether session depth (page_loads, searches_count, total feature interactions) differs by device type. If mobile sessions are shallower (fewer page loads, fewer feature uses), it suggests a need for optimised mobile UX for the risk-check workflow.

⚙️ Model / Method

Mann-Whitney U test: session_depth (Σ all feature counts) by device_type; Mann-Whitney on session_duration = last_seen - first_seen; box plot comparison.

📰 Computers in Human Behavior · International Journal of Human-Computer Studies

Crisis Informatics / Platform Research

Emergency Panel Opens as a Crisis Signal

📋 Example

The emergency_opens column captures when users activated the platform's emergency protocol during their session. A researcher extracts sessions where emergency_opens > 0, notes their timestamps, and cross-references with incident dates in the archive — testing whether emergency panel activations cluster on days with high-severity incidents.

⚙️ Model / Method

Filter sessions where emergency_opens > 0; aggregate daily count of emergency sessions; cross-correlate with daily incident count from archive (Pearson/Spearman); note session count with emergency_opens > 0 as denominator for prevalence.

📰 Disasters · International Journal of Disaster Risk Reduction

Platform Analytics / User Behaviour

Session Temporal Patterns: Time of Day and Day of Week

📋 Example

The first_seen timestamp allows analysis of when users access the platform. A researcher extracts hour-of-day and day-of-week from first_seen, producing a heatmap of session starts by hour × day — revealing whether the platform is primarily used during morning planning periods (before travel) or throughout the day.

⚙️ Model / Method

Parse first_seen to hour and day_of_week; contingency table: hour × day_of_week × device_type; chi-square test of hour distribution by device_type (do mobile users access at different hours?).

📰 MIS Quarterly · Decision Support Systems

📊

API Traffic Log — 5,096 Records

5,096 records · 4 use cases · CSV · PII (IP, user_agent) redacted on download

Platform Usage & Demand

Platform Analytics / System Design

Endpoint Usage Distribution and Platform Feature Demand

📋 Example

The endpoint column reveals which platform features receive the most calls: api_feed dominates at 3,203 requests (63%), followed by index page loads (1,133), api_risk (475) and routing endpoints (api_route: 59, api_route_visual: 45, route_recommend: 44, api_safe_route: 11). A researcher builds a usage pyramid and tests whether this matches the searches dataset pattern — both show risk queries dominating over route queries.

⚙️ Model / Method

Frequency table of endpoint; proportion per endpoint; compare endpoint proportions with action_taken proportions from searches dataset (chi-square test of proportional equivalence).

📰 Information Systems Research · MIS Quarterly

Systems Performance / Platform Engineering

API Response Time Analysis by Endpoint

📋 Example

The response_time_ms column enables latency analysis per endpoint. A researcher computes p50, p95 and p99 response times for each endpoint — testing whether AI-heavy endpoints (api_risk, which calls DeepSeek) have systematically higher and more variable response times than static endpoints (index, api_feed cache hits).

⚙️ Model / Method

Compute percentiles (p50, p95, p99) of response_time_ms per endpoint; box plot per endpoint; Mann-Whitney U test: api_risk response time vs. api_feed response time; flag requests where response_time_ms > 5,000 ms as slow.

📰 Future Generation Computer Systems · Journal of Systems and Software

Platform Geography / Demand Mapping

Spatial Distribution of API Requests

📋 Example

Plotting 5,096 request origins by lat/lng reveals the geographic demand surface for the platform. A researcher applies KDE to identify demand clusters — is platform usage concentrated in Lagos, Abuja and Port Harcourt, or does it extend to smaller cities and rural areas? This maps the real geographic footprint of SafeZone.ng usage.

⚙️ Model / Method

KDE on (lat, lng) with Silverman bandwidth; overlay on Nigerian state boundaries; compute requests per state by assigning each (lat,lng) to its state polygon; Gini coefficient of geographic concentration.

📰 Journal of Economic Geography · Urban Studies

Platform Reliability / Information Systems

HTTP Error Rate and Platform Reliability

📋 Example

The response_code column records HTTP status for each request. A researcher computes the error rate: what proportion of 5,096 requests returned 4xx or 5xx codes? Error rates by endpoint reveal whether specific features are less reliable — e.g. do routing endpoints (which call external Google APIs) have higher error rates than cached feed endpoints?

⚙️ Model / Method

Binary: success (2xx) vs. error (4xx/5xx); error rate per endpoint; chi-square test of error rate differences across endpoints; time series of daily error count.

📰 Information Systems Research · IEEE Transactions on Services Computing

📡

Live Intelligence Feed — 102 Records

102 records · 4 use cases · CSV

Platform Usage & Demand

Platform Content Analysis / Information Science

Feed Content Composition: Incident Types and Satellite Data Mix

📋 Example

A researcher analyses the type column to document what categories of intelligence the live feed delivers. The mix of structured incident types (kidnapping, terrorism, arrest) alongside raw NASA FIRMS satellite text strings reveals a data quality issue: some type fields contain full prose rather than coded categories. The researcher quantifies the proportion of well-structured vs. unstructured type entries.

⚙️ Model / Method

Frequency table of type values; classify each type as "coded" (short category word) vs. "unstructured" (long NASA text); proportion of coded entries = data quality metric; identify which source column values correlate with unstructured type fields.

📰 Information Systems Research · Data & Knowledge Engineering

Information Retrieval / Platform Design

Distance-to-User Relevance: How Close Are Feed Items to Queried Locations?

📋 Example

The distance_meters column captures how far each feed item is from the user's queried location when the feed was served. A researcher analyses the distribution of distance_meters: what is the median, p75, p95 distance? How often does the feed serve items more than 100 km from the user's location? This directly informs feed relevance algorithm design.

⚙️ Model / Method

Histogram of distance_meters; compute percentiles; stratify by severity: do HIGH severity items appear closer to the queried location (intentional prioritisation) or is distance independent of severity?

📰 Information Systems Research · ACM Transactions on Information Systems

Intelligence Quality / Information Science

Confidence Score Distribution Across Feed Item Types

📋 Example

The confidence column rates each feed item. A researcher compares confidence distributions across incident categories — are arrests systematically higher-confidence (official sources) than communal_violence events? Are NASA FIRMS satellite detections assigned a specific confidence value or null? This documents the platform's confidence scoring behaviour per source type.

⚙️ Model / Method

Box plot of confidence by type (for coded type values); Mann-Whitney test between high-confidence source types (arrest, terrorism) and lower-confidence types (communal_violence, other); tabulate confidence for NASA FIRMS entries.

📰 Intelligence and National Security · Journal of Information Science

Conflict Epidemiology / Public Health

Casualty Reporting in the Live Feed: killed, wounded, abducted

📋 Example

The killed, wounded and abducted columns provide casualty counts per feed item. A researcher computes total casualties reported across 102 feed items, their distribution by incident type, and the proportion of items with zero vs. non-zero casualties. Cross-referencing with the confidence column shows whether higher-confidence items also report more specific casualty figures.

⚙️ Model / Method

Sum and mean of killed + wounded + abducted overall and by type; proportion of items with casualties > 0; Spearman correlation between confidence and casualties > 0 (binary); compare to archive dataset casualty totals for consistency.

📰 The Lancet Regional Health – Africa · Journal of Conflict Resolution

📍

Geocoding Cache — 422 Records

422 records · 4 use cases · CSV

Geospatial Intelligence

GIS / Platform Design Research

Geocoding Source Usage: Nominatim vs. Google

📋 Example

The source column identifies whether each of the 422 places was resolved via OpenStreetMap Nominatim or Google Geocoding API. A researcher computes the split and examines whether certain place name types (state names, LGA names, street-level queries) are routed to one provider vs. the other — documenting the platform's geocoding fallback logic.

⚙️ Model / Method

Frequency table: source × resolved vs. null; characterise place name length and type (single word vs. multi-word) by source; test whether source differs by place name category.

📰 International Journal of Geographical Information Science · Transactions in GIS

Computational Linguistics / GIS

User Place Name Variation: Raw vs. Canonical Name Comparison

📋 Example

Comparing the place column (raw user input) with resolved_name (canonical output) reveals toponym variations: "niger state" vs. "Niger State", "kogi" vs. "Kogi", abbreviated or informal names. A researcher catalogues these variations, computes the edit distance between raw and resolved name, and identifies place name pairs where the platform resolved correctly vs. incorrectly.

⚙️ Model / Method

Levenshtein distance for each (place, resolved_name) pair; categorise into exact match, case-only difference, truncation, common abbreviation, ambiguous; manual review of high-distance pairs for error analysis.

📰 International Journal of Geographical Information Science · Natural Language Engineering

Platform Analytics / Geographic Demand

Cache Hit Frequency: Which Locations Are Most Queried

📋 Example

Some place names appear multiple times in the cache (repeated queries). A researcher counts the frequency of each resolved_name to identify the most-queried locations — do state names (Nigeria, Lagos, Abuja) dominate, or do specific neighbourhood names (Shomolu, Ikoyi, Ikeja) appear frequently? This reveals where platform users are actually located when querying.

⚙️ Model / Method

Value_counts of resolved_name; top-20 most-cached locations; map top-50 by frequency; compare to state-level query distribution in searches dataset.

📰 Urban Studies · Journal of Economic Geography

Data Quality / GIS Research

Geocoding Resolution Quality: Coordinate Plausibility Check

📋 Example

For each of the 422 resolved coordinates, a researcher checks whether the lat/lng falls within Nigeria's bounding box (3.0°E to 15.0°E, 4.0°N to 14.0°N). Any coordinate outside this box represents a geocoding error or a legitimate cross-border query. This basic plausibility check documents the resolution error rate and identifies problematic place names.

⚙️ Model / Method

Binary flag: in_bounding_box = (3 ≤ lng ≤ 15) AND (4 ≤ lat ≤ 14); count out-of-box entries; for out-of-box entries, identify the place name and what the platform resolved it to; resolution error rate = out-of-box / 422.

📰 International Journal of Geographical Information Science · Data & Knowledge Engineering

⛰️

Elevation Cache — 9 Records

9 records · 4 use cases · CSV with embedded JSON payload

Geospatial Intelligence

Transport Geography / Route Engineering

Elevation Profile Extraction and Route Gradient Analysis

📋 Example

⚙️ Model / Method

📰 Transportation Research Part D · Journal of Transport Geography

GIS Data Quality / Research Methodology

Elevation Data Resolution and Its Implications for Route Accuracy

📋 Example

⚙️ Model / Method

📰 International Journal of Geographical Information Science · Transactions in GIS

Platform Geography / GIS

Platform Elevation Coverage: Where Does SafeZone.ng Profile Terrain?

📋 Example

⚙️ Model / Method

📰 Journal of Transport Geography · Transactions in GIS

Research Methodology / Data Provenance

Data Source Documentation for Elevation Intelligence

📋 Example

⚙️ Model / Method

Extract data_source from all 9 payloads; frequency table; cross-reference provider accuracy specifications with known RMSE for Nigeria from published DEM accuracy studies.

📰 International Journal of Geographical Information Science · Remote Sensing

🛰️

NASA FIRMS Fire Clusters — 1 Record

1 records · 4 use cases · CSV · NASA FIRMS VIIRS-sourced

Geospatial Intelligence

Platform Design / Remote Sensing Integration

NASA FIRMS Integration: How SafeZone.ng Incorporates Satellite Fire Data

📋 Example

⚙️ Model / Method

📰 Remote Sensing · Computers & Geosciences

Remote Sensing / Environmental Monitoring

Fire Radiative Power as Indicator of Fire Intensity

📋 Example

⚙️ Model / Method

📰 Remote Sensing of Environment · International Journal of Wildland Fire

Environmental Geography / Land Use

Geographic Context: Northern Nigeria / Lake Chad Fire Detection

📋 Example

⚙️ Model / Method

📰 Journal of Arid Environments · Remote Sensing

Research Methodology / Data Infrastructure

Dataset Limitation Documentation: Building a Larger Fire Intelligence Dataset

📋 Example

⚙️ Model / Method

📰 Remote Sensing · Data in Brief

⛺

IDP Camps — 10 Records

10 records · 4 use cases · CSV

Humanitarian & Social Context

Humanitarian Studies / Public Administration

Camp Capacity Utilisation and Overcrowding Assessment

📋 Example

For each of the 10 camps, capacity and current_population are both recorded. A researcher computes utilisation rate = current_population / capacity and identifies camps that are over-capacity (utilisation > 1.0). The Zamfara camp (pop. 65,000) is particularly relevant given banditry displacement context. Overcrowding has direct implications for managing agency resource planning.

⚙️ Model / Method

Utilisation rate per camp; rank by utilisation; flag over-capacity camps; cross-reference managing_agency with the 10 camps to identify which agencies manage the most overcrowded sites.

📰 Forced Migration Review · Journal of Refugee Studies

Human Geography / Conflict Studies

Geographic Distribution of Displacement: State-Level Pattern

📋 Example

Plotting the 10 camps by lat/lng, a researcher maps the geographic distribution of displacement — 5 of 10 camps are in Borno (Northeast conflict zone), with others in Benue, Plateau (Middle Belt communal conflict), Zamfara (Northwest banditry) and Taraba. This spatial pattern reflects distinct conflict drivers in different regions of Nigeria.

⚙️ Model / Method

Point map of 10 camps coloured by state; aggregate current_population by state; bar chart: population per state; contextualise against incident archive data by state to link displacement to recorded events.

📰 Political Geography · African Affairs

Humanitarian Governance / Aid Effectiveness

Agency Co-ordination: Multi-Agency vs. Single-Agency Camp Management

📋 Example

The managing_agency field records one or more agencies (e.g. "UNHCR/NEMA", "SEMA/NRC"). A researcher counts agencies per camp and identifies which camps have multi-agency co-ordination vs. single-agency management. The Makurdi site (pop. 47,000) managed by NEMA/Benue SEMA is compared to the smaller Banki Transit Camp (6,200, UNHCR/IRC) to assess resource allocation relative to population.

⚙️ Model / Method

Count agencies per camp (split managing_agency by "/"); binary: single vs. multi-agency; cross-tabulate: multi-agency × above/below median current_population; note: n=10 does not support regression.

📰 World Development · Disasters

Conflict Geography / Humanitarian Security

Proximity of IDP Camps to Security Incidents

📋 Example

Using the 10 camp coordinates and the 1,774 incident records, a researcher computes the distance from each camp to the nearest recorded incident, and counts incidents within 50 km of each camp. Camps in Borno (where 5 of 10 are located) likely have the highest nearby incident counts. This quantifies the security environment surrounding each displacement site.

⚙️ Model / Method

Haversine distance from each camp to each incident; nearest incident distance per camp; count incidents within 50 km buffer per camp; filter by severity: how many HIGH/critical incidents are within 50 km?

📰 Disasters · Journal of Humanitarian Logistics and Supply Chain Management

✊

Labour Strikes & Disruptions — 6 Records

6 records · 4 use cases · CSV

Humanitarian & Social Context

Supply Chain Risk / Labour Economics

Logistics Impact Classification of Nigerian Labour Disruptions

📋 Example

⚙️ Model / Method

📰 Industrial Relations · Supply Chain Management: An International Journal

Supply Chain Disruption / Case Study Research

NARTO Road Transport Strike: Documented Supply Chain Impact Case Study

📋 Example

⚙️ Model / Method

📰 Supply Chain Management: An International Journal · Journal of Operations Management

Labour Economics / Industrial Relations

Sector-Specific Strike Frequency and Duration Patterns

📋 Example

⚙️ Model / Method

📰 British Journal of Industrial Relations · Journal of African Economies

Information Science / Platform Design

Platform Strike Monitoring: How the SafeZone.ng Dataset Was Built

📋 Example

⚙️ Model / Method

📰 Information Systems Research · Computers & Industrial Engineering

🗳️

Elections Calendar — 6 Records

6 records · 4 use cases · CSV

Humanitarian & Social Context

Political Science / Electoral Studies

Election Risk Level Documentation and Context

📋 Example

⚙️ Model / Method

📰 Electoral Studies · Journal of Modern African Studies

Logistics Planning / Political Risk Management

Election Calendar as a Logistics Planning Tool

📋 Example

⚙️ Model / Method

📰 International Journal of Logistics Management · Journal of Business Logistics

Quantitative Political Science / Security Research

Incident Archive Cross-Reference: Do Past Elections Show Incident Spikes?

📋 Example

⚙️ Model / Method

📰 Electoral Studies · African Affairs

Political Risk Research / Country Risk

2027 General Election: Nigeria's Highest-Risk Electoral Event

📋 Example

⚙️ Model / Method

📰 Journal of Modern African Studies · African Affairs

🎓

Universities — 55 Records

55 records · 4 use cases · CSV

Humanitarian & Social Context

Education Policy / Regional Development

University Type Distribution Across Nigerian Geopolitical Zones

📋 Example

With type (Federal/State/Private), state and lat/lng, a researcher maps the geographic distribution of universities across Nigeria's six geopolitical zones. The analysis tests whether Federal universities are more equitably distributed than State or Private ones — and whether Private university growth has concentrated in specific states (historically South-West heavy) while other zones rely on State institutions.

⚙️ Model / Method

Count universities by type × geopolitical zone; Location Quotient of each type per zone relative to zone population; Theil decomposition of geographic concentration by type.

📰 Higher Education Policy · Journal of African Economies

Campus Security / Educational Geography

Incident Proximity: Which Nigerian Universities Are in High-Risk Areas?

📋 Example

Using the 55 university coordinates and the 1,774 incident records, a researcher computes the number and severity of incidents within a 25 km radius of each university campus. Universities in Borno, Zamfara, Kaduna and Plateau are expected to have more nearby incidents than Lagos or Ogun campuses — quantifying the campus security environment for each of the 55 institutions.

⚙️ Model / Method

Haversine distance from each university to each incident; count incidents within 25 km; severity-weighted score per university; rank 55 universities from most- to least-proximate to recorded incidents; produce a campus risk tier map.

📰 Higher Education · Journal of Education Policy

Higher Education Research / Demography

Student Population Distribution and Institutional Scale

📋 Example

The students_approx column provides enrolment estimates. A researcher analyses the distribution of enrolment across the 55 institutions: what is the range (smallest to largest)? What is the total approximate student population? Does enrolment correlate with institution type (Federal universities larger than State which are larger than Private)?

⚙️ Model / Method

Histogram of students_approx; total = Σ(students_approx) across 55; Kruskal-Wallis test: enrolment distribution by type (Federal/State/Private); note these are approximate figures, not official NUC statistics.

📰 Higher Education · Oxford Review of Education

Education History / Policy Research

University Founding Year and Historical Development of Nigerian Higher Education

📋 Example

The founded column spans from Nigeria's oldest universities (1948, 1962) to more recent Private institutions. A researcher plots a timeline of university founding dates, identifying: the first wave (pre-independence, 1948–1960), the federal expansion era (1970s–1980s), the state university proliferation (1990s) and the private university boom (2000s–present).

⚙️ Model / Method

Histogram of founded year by decade; stacked by type; Mann-Whitney test: founding year by type (are Private universities newer?); overlay with Nigeria education policy milestones.

📰 History of Education · Higher Education Policy

📐 Column Reference

Dataset Schemas

Exact column names from each source file. PII columns (ip, session_id, user_agent) are redacted before download.

🛡️

Security Incidents

Corridor Security Intelligence

Column	Description
`id`	Unique incident ID
`lat`	Latitude
`lng`	Longitude
`type`	Incident type (kidnapping, armed_robbery, banditry, terrorism, cult_violence, etc.)
`severity`	Severity (low / medium / high / critical)
`title`	Incident headline
`description`	Full incident description
`source`	Source media outlet
`date_occurred`	Date incident occurred
`date_added`	Date added to platform
`country`	Country
`region`	Region
`verified`	Verified flag (0/1)

📁

Incident Archive — 38 Daily Snapshots

Corridor Security Intelligence

Column	Description
`id`	Incident ID
`title`	Headline
`type`	Incident type
`category`	Category
`severity`	Severity
`lat`	Latitude
`lng`	Longitude
`state`	State
`location_name`	Location name
`killed`	Fatalities count
`wounded`	Wounded count
`abducted`	Abducted count
`source`	Source outlet
`source_url`	Source URL
`pub_time`	Event publication time
`confidence`	Confidence score
`country`	Country
`archived_at`	Timestamp this record was archived

🔍

Intelligence-Grade Incidents — 187 Records

Corridor Security Intelligence

Column	Description
`id`	Incident ID
`title`	Headline
`description`	Full description
`type`	Incident type
`category`	Category
`severity`	Severity
`lat`	Latitude
`lng`	Longitude
`state`	State
`lga`	LGA
`location_name`	Location name
`killed`	Fatalities
`wounded`	Wounded
`abducted`	Abducted
`source`	Source outlet
`source_url`	Source URL
`pub_time`	Publication time
`confidence`	Confidence score
`country`	Country
`verified`	Verified (0/1)
`feed_ts`	Feed timestamp
`harvest_run`	Harvest run ID
`segment_id`	Road segment ID (route network)
`corridor_id`	Corridor ID (route network)
`snap_confidence`	Snap confidence to road network

⚓

Maritime Security Incidents — 6 Records

Corridor Security Intelligence

Column	Description
`id`	Incident ID
`date`	Date
`lat`	Latitude
`lng`	Longitude
`area`	Sea area
`vessel_name`	Vessel name
`vessel_type`	Vessel type
`flag`	Flag state
`attack_type`	Attack type
`attack_description`	Attack description
`crew_outcome`	Crew outcome
`crew_kidnapped`	Crew kidnapped (count)
`crew_assaulted`	Crew assaulted (count)
`outcome`	Incident overall outcome
`perpetrators`	Perpetrators described
`distance_from_shore_nm`	Distance from shore (nautical miles)
`reporting_authority`	Reporting authority
`source`	Source outlet
`source_url`	Source URL
`confidence`	Confidence score
`logged_at`	Date logged

👥

Armed Groups Register — 8 Actors

Corridor Security Intelligence

Column	Description
`id`	Actor ID
`name`	Group name
`aliases`	Known aliases (pipe-separated)
`type`	Group type
`ideology`	Ideology
`origin_state`	State of origin
`primary_states`	Primary operational states (pipe-separated)
`tactics`	Tactics used (pipe-separated)
`primary_targets`	Primary targets (pipe-separated)
`estimated_strength`	Estimated strength
`activity_status`	Activity status
`last_confirmed_date`	Last confirmed active date
`last_confirmed_location`	Last confirmed location
`known_commanders`	Known commanders
`known_affiliates`	Known affiliates
`notable_incidents`	Notable incidents
`threat_level`	Threat level
`source_refs`	Source references
`created_at`	Record created
`updated_at`	Record updated
`notes`	Notes

🗺️

Platform Search Queries — 1,508 Records

Route Intelligence & Network Data

Column	Description
`id`	Search ID
`session_id`	Session ID (redacted in download)
`timestamp`	Query timestamp
`search_type`	Search type: risk / route / hotel / university
`search_term`	Raw text typed by user
`resolved_name`	Resolved location name
`lat`	Latitude of resolved location
`lng`	Longitude of resolved location
`state`	Nigerian state of resolved location
`country`	Country of session origin
`result_count`	Number of results returned
`action_taken`	Action platform took: risk_check / route_check / route_recommend / safe_route / search / safety
`device_type`	Device type: mobile / desktop
`month_year`	Month-year bucket (YYYY-MM)

🛣️

Road Segment Cache — 64 Records

Route Intelligence & Network Data

Column	Description
`cache_key`	Cache key (lat,lng string)
`ts`	Unix timestamp of when cached
`payload`	JSON payload: segment_id, segment_name, corridor_id, snap_distance_m, snap_confidence, snap_lat, snap_lng, is_choke_point, alt_routes, road_class, state, lga

📐

Distance Matrix Cache — 6 Records

Route Intelligence & Network Data

Column	Description
`cache_key`	Cache key (hash)
`data`	JSON payload: origin_addresses, destination_addresses, rows[{status, distance_m, distance_text, duration_s, duration_text, duration_traffic_s, duration_traffic_text}]
`expires_at`	Cache expiry timestamp

🏨

Hotels & Logistics Hubs — 304 Records

Logistics Infrastructure & Facilities

Column	Description
`id`	Hotel ID
`name`	Full name
`short_name`	Short name
`city`	City
`state`	State
`lat`	Latitude
`lng`	Longitude
`stars`	Star rating (1–5)
`type`	Facility type (luxury/budget/guesthouse/etc.)
`address`	Physical address
`phone`	Phone number
`price_range`	Price range (budget/mid/high)
`chain`	Chain affiliation (or null)
`verified_coords`	Coordinate verified flag
`osm_id`	OpenStreetMap ID

🏗️

Critical Infrastructure Assets — 12 Records

Logistics Infrastructure & Facilities

Column	Description
`id`	Asset ID
`name`	Asset name
`asset_type`	Type (crude_pipeline/gas_pipeline/rail/port/power_generation/power_transmission/telecoms/bridge/lng_terminal)
`operator`	Operator(s)
`corridor`	Corridor affiliation
`lat`	Latitude
`lng`	Longitude
`state`	State
`lga`	LGA
`description`	Description
`strategic_importance`	Strategic importance (CRITICAL or HIGH)
`last_attack_date`	Date of last recorded attack
`attack_count`	Total recorded attacks
`status`	Operational status
`created_at`	Record created
`notes`	Notes

📱

User Sessions — 1,649 Records

Platform Usage & Demand

Column	Description
`session_id`	Session ID (redacted in download)
`device_type`	Device type: desktop (1,045) / mobile (604)
`first_seen`	First platform access timestamp
`last_seen`	Last platform access timestamp
`page_loads`	Total page loads in session
`searches_count`	Searches performed
`risk_checks`	Risk assessments run
`route_checks`	Route checks performed
`hotel_checks`	Hotel lookups performed
`uni_checks`	University lookups performed
`emergency_opens`	Emergency panel opens
`fact_checks`	Fact checks performed
`country`	Country (all sessions: Nigeria)
`referrer`	Referrer URL

📊

API Traffic Log — 5,096 Records

Platform Usage & Demand

Column	Description
`id`	Request ID
`lat`	Request origin latitude
`lng`	Request origin longitude
`timestamp`	Request timestamp
`endpoint`	API endpoint: api_feed/index/api_risk/api_route/api_route_visual/route_recommend/api_route_cached/api_safe_route/api_safe_route_cached/api_university/admin_login
`response_code`	HTTP response code
`response_time_ms`	Response time (milliseconds)

📡

Live Intelligence Feed — 102 Records

Platform Usage & Demand

Column	Description
`id`	Feed item ID
`title`	Headline
`type`	Type: incident category or NASA FIRMS satellite text
`category`	Category
`severity`	Severity
`lat`	Latitude
`lng`	Longitude
`state`	State
`location_name`	Location name
`killed`	Fatalities
`wounded`	Wounded
`abducted`	Abducted
`source`	Source outlet
`source_url`	Source URL
`pub_time`	Publication time
`confidence`	Confidence score
`distance_meters`	Distance from user's queried location (metres)
`feed_ts`	Feed delivery timestamp

📍

Geocoding Cache — 422 Records

Geospatial Intelligence

Column	Description
`place`	Raw place name as typed by user
`lat`	Resolved latitude
`lng`	Resolved longitude
`resolved_name`	Canonical resolved place name
`source`	Geocoding source: nominatim (OSM) or Google
`cached_at`	Geocoding timestamp

⛰️

Elevation Cache — 9 Records

Geospatial Intelligence

Column	Description
`cache_key`	Cache key
`data`	JSON payload: profile[{lat, lng, elevation_m, resolution_m}], summary, data_source, cached_at, cache_key
`cached_at`	Cache timestamp

🛰️

NASA FIRMS Fire Clusters — 1 Record

Geospatial Intelligence

Column	Description
`cluster_id`	Cluster ID
`lat`	Latitude (13.07152)
`lng`	Longitude (14.78472)
`state`	State (not populated for this record)
`lga`	LGA (not populated for this record)
`fire_count`	Number of fire detections (6)
`max_frp`	Max fire radiative power (6.69 MW)
`avg_frp`	Average FRP (6.2 MW)
`radius_km`	Cluster radius (15 km)
`detected_date`	Detection date (2026-06-29)
`confidence`	Confidence (LOW)
`is_bandit_context`	Bandit attack context flag (0)
`agricultural_burn_possible`	Agricultural burn flag (0)
`cluster_title`	Cluster title
`cluster_desc`	Cluster description
`ds_interpretation`	AI interpretation
`ds_actor`	Actor attribution
`ds_anomaly_score`	Anomaly score
`created_at`	Created timestamp

⛺

IDP Camps — 10 Records

Humanitarian & Social Context

Column	Description
`id`	Camp ID
`name`	Camp name
`lat`	Latitude
`lng`	Longitude
`state`	State
`lga`	LGA
`camp_type`	Camp type (urban_camp/rural_settlement/transit)
`status`	Status (all active)
`capacity`	Designed capacity
`current_population`	Current population
`managing_agency`	Managing agency
`open_date`	Open date
`last_verified`	Last verification date
`security_level`	Security level
`access_status`	Access status
`notes`	Notes

✊

Labour Strikes & Disruptions — 6 Records

Humanitarian & Social Context

Column	Description
`id`	Strike ID
`date_start`	Start date
`date_end`	End date
`union`	Union or industrial body
`sector`	Sector
`states_affected`	States affected
`description`	Description
`impact_on_logistics`	Documented logistics impact (as assessed)
`status`	Status (resolved/ongoing)
`source`	Source
`logged_at`	Date logged

🗳️

Elections Calendar — 6 Records

Humanitarian & Social Context

Column	Description
`id`	Election ID
`election_type`	Election type
`date`	Scheduled date
`state`	State (or "National")
`lga`	LGA scope
`status`	Status (scheduled)
`candidates_count`	Number of registered candidates
`issues`	Security and logistics context notes
`risk_level`	Risk level (CRITICAL/HIGH/MEDIUM)
`notes`	Additional notes
`created_at`	Record created

🎓

Universities — 55 Records

Humanitarian & Social Context

Column	Description
`id`	University ID
`name`	Full institution name
`short_name`	Common short name
`state`	State
`city`	City
`lat`	Latitude
`lng`	Longitude
`type`	Type: Federal / State / Private
`founded`	Year founded
`students_approx`	Approximate student enrolment
`website`	Website URL

SafeZone.ng Research Datasets

Research Datasets

Corridor Security Intelligence

Route Intelligence & Network Data

Logistics Infrastructure & Facilities

Platform Usage & Demand

Geospatial Intelligence

Humanitarian & Social Context

Research Use Cases — 80 Studies

Security Incidents

Incident Archive — 38 Daily Snapshots

Intelligence-Grade Incidents — 187 Records

Maritime Security Incidents — 6 Records

Armed Groups Register — 8 Actors

Platform Search Queries — 1,508 Records

Road Segment Cache — 64 Records

Distance Matrix Cache — 6 Records

Hotels & Logistics Hubs — 304 Records

Critical Infrastructure Assets — 12 Records

User Sessions — 1,649 Records

API Traffic Log — 5,096 Records

Live Intelligence Feed — 102 Records

Geocoding Cache — 422 Records

Elevation Cache — 9 Records

NASA FIRMS Fire Clusters — 1 Record

IDP Camps — 10 Records

Labour Strikes & Disruptions — 6 Records

Elections Calendar — 6 Records

Universities — 55 Records

Dataset Schemas

Security Incidents

Incident Archive — 38 Daily Snapshots

Intelligence-Grade Incidents — 187 Records

Maritime Security Incidents — 6 Records

Armed Groups Register — 8 Actors

Platform Search Queries — 1,508 Records

Road Segment Cache — 64 Records

Distance Matrix Cache — 6 Records

Hotels & Logistics Hubs — 304 Records

Critical Infrastructure Assets — 12 Records

User Sessions — 1,649 Records

API Traffic Log — 5,096 Records

Live Intelligence Feed — 102 Records

Geocoding Cache — 422 Records

Elevation Cache — 9 Records

NASA FIRMS Fire Clusters — 1 Record

IDP Camps — 10 Records

Labour Strikes & Disruptions — 6 Records

Elections Calendar — 6 Records

Universities — 55 Records

How to Cite This Dataset

Request Dataset