Slymar19/Kwara-State-Election-Chronicles

# Kwara State Election Chronicles: Spatial Analysis of the 2023 Nigerian Presidential Election **View Interactive Dashboard →** github.com *(20-30 second screen recording showing filter interactivity and radius selection)* --- ## Overview This project applies spatial statistics and machine learning to 2,518 polling units across Kwara State, Nigeria, to identify polling units with statistically unusual voting patterns in the 2023 presidential election — potential indicators warranting further scrutiny. Historical trends from 2015–2023 provide additional context on the state's shifting political landscape. **Methods used:** - **DBSCAN** — geographic clustering to identify spatially isolated polling units, run at three radii (500m, 1km, 2km) - **Local Moran's I** — spatial autocorrelation of turnout rate, identifying statistically significant clusters of unusually high or low turnout relative to neighboring units - **Getis-Ord Gi\*** — hotspot/coldspot analysis on party vote counts - **Isolation Forest** — multivariate anomaly detection on vote-count and turnout features - **Composite outlier score** — a normalized combination of the above, used to rank the highest-risk polling units ## Key Findings - Local Moran's I (run on turnout rate) identified 986 polling units as statistically significant spatial clusters (p < 0.05) - Cross-referencing against Getis-Ord Gi* flags narrowed this to 288 High-High and 245 Low-Low locations — confirmed by two independent methods - DBSCAN flagged 471 / 329 / 243 polling units as spatial outliers at 500m / 1km / 2km respectively - The composite outlier score identifies Ajikobi (Ilorin West LGA) and TSsonga 111 (Edu LGA) as the highest-risk areas for further investigation - Labour Party's state-level vote share rose from ~0.01% (2019) to 7.71% (2023), coinciding with Peter Obi's candidacy - APC–PDP vote margins tightened in 6 of 9 tracked LGAs between 2019 an …