charles-bruce/Violence-Analysis

This project explores violence in Nigeria. It identifies factors linked to high fatalities (e.g., event types, actors) via data exploration and visualizations. It also builds Logistic Regression and Random Forest models to predict violence escalation, optimizing performance through feature engineering and tuning. # Exploring Factors of Violence in Nigeria and Predicting Escalation This project investigates conflict-related fatalities in Nigeria using two datasets: **Nigeria_ACLED.csv** and **violence.csv**. The goal is to identify factors associated with high fatalities and build predictive models for violence escalation. ## Part I: Data Exploration The first part of this project explores the **Nigeria_ACLED.csv** dataset, focusing on the variable *‘ACLED Fatalities’*. - **Objective**: Identify factors most associated with high fatalities. - **Methodology**: - Analyzed key variables such as event types, actors, geographic locations, and ACLED notes. - Applied data cleaning, feature engineering (e.g., binning and categorization), and statistical analysis. - Visualized trends and associations using clear, interpretable plots. - **Findings**: Key factors contributing to high fatalities were identified and discussed, providing insights into patterns of violence in Nigeria. ## Part II: Machine Learning The second part uses the **violence.csv** dataset to predict the *‘Class’* variable, which indicates violence escalation. - **Objective**: Build predictive models to classify violence escalation. - **Methodology**: 1. Selected an evaluation metric based on dataset characteristics and project goals. 2. Trained and evaluated two models: - Logistic Regression - Random Forest 3. Compared model performance to determine the best fit for the task. 4. Applied techniques to improve performance (e.g., feature selection, hyperparameter tuning). - **Results**: Performance of both models was evaluated and enhanced, with explanations of the methods used for improvement. ## Tools & Technologies - Python - Pandas, NumPy, Matplotlib, Seaborn - Scikit-learn - Geo-spatial and NLP techniques (where applicable)