imbukwa1/AI-_driven_pastrolist-conflict-warning-system-

AI-Driven Predictive Analytics for Pastoralist Conflict Prevention in Arid Kenya Pastoralist Resource Stress & Conflict Early-Warning System Project Overview This project develops a data-driven early-warning system for detecting pasture stress and potential conflict hotspots in arid and semi-arid pastoralist regions of Kenya, with a focus on Wajir County (and extendable to Turkana, Marsabit, Samburu, and West Pokot). The system integrates satellite-derived environmental indicators with simulated mobility (CDR) anomaly detection to flag emerging risk zones before crises escalate. Results are visualized through an interactive Streamlit dashboard to support early interventions such as water trucking, fodder deployment, livestock offtake, and peace actions. Objectives Detect vegetation and rainfall stress using NDVI and CHIRPS data Forecast future pasture stress hotspots at 5 km spatial resolution Identify abnormal livestock/human movement patterns using anomaly detection Provide explainable AI outputs (SHAP) to build trust with stakeholders Support early conflict prevention and climate resilience planning Data Sources Environmental Data NDVI (MODIS) – vegetation greenness CHIRPS rainfall – 16-day accumulated precipitation 5 km spatial grid points – uniform spatial analysis Mobility Data (Simulated) Anonymized Call Detail Record (CDR)–like mobility data Used to represent livestock and pastoralist movement patterns Aggregated at grid-cell level (no personal data) No personal or sensitive data is used in this project. Methodology 1️ Environmental Stress Modeling NDVI anomalies computed using seasonal climatology Rainfall deficits extracted per grid cell Lagged features added for temporal learning Machine learning model (Logistic Regression / XGBoost baseline) Output: Probability of future pasture stress hotspot 2️ Anomaly Detection (Recommendation #2) Simulated CDR mobility features: Movement intensity Directional changes Spatial congregation Isolation Forest used to detect abnormal movement patterns High anomaly scores i …