Endalkmin/conflict-event-impact-classifier

Predicting severity of impact from conflict events in Kenya. # Predicting Severity of Impact from Conflict Events in Kenya. ## Problem Statement Using historical data from the Armed Conflict Location & Event Data research (ACLED), this research develops a predictive model to categorise conflict incidents in Kenya into four severity levels (Low, Moderate, High, and Critical). By achieving the accuracy, precision, F1-score and confusion matrix the approach assists government agencies, peacekeeping forces and humanitarian organisations in better allocating resources and prioritising response activities. ## 1. Business Understanding Kenya has had a lot of protests in recent years, motivated by social, political, and economic issues. These incidents frequently intensify quickly and have different effects on the community. Decision-makers do not currently have access to efficient real-time tools for anticipating possible outcomes. Objectives : a) Establish a predictive model that divides conflict incidents into four tiers of severity. b) Using past ACLED data, to create a Community Impact Score (CIS). c) Assist stakeholders in allocating resources and setting priorities for response activities. ## 2. Data Understanding The project makes use of information from the Armed Conflict Location & Event Data initiative (ACLED), which offers comprehensive logs of events connected to conflicts throughout Africa. More than 431,000 records from 1997 to 2025 are included in the raw dataset, with 17,812 events occurring in Kenya alone. Significant factors consist of: - Fatalities. - Actors Involved. - Event type (Protests, Battles, Explosions/Remote violence, etc.). - Civilian targeting. - Narrative descriptions. - Geographic coordinates. ## 3. Data Wrangling Here we will work on : - Filtering Kenya-specific data (17,812 events). - Handling missing values (imputation and column dropping). - Text data cleaning (removing special characters, lowercasing). - Feature engineering: -Combined actor columns. -Created Community Impact Score …