eleni-abebe/Ethiopia-road-accident-prediction

🚦 Ethiopia Road Accident Severity Prediction A machine learning–based system for predicting road accident severity in Addis Ababa, Ethiopia, using historical road traffic accident data. The project includes data preprocessing, model training, evaluation, and an interactive Streamlit web application. 📌 Project Overview Road traffic accidents are a major public safety issue in Ethiopia. This project aims to predict the severity of road accidents (`Fatal`, `Serious`, `Slight`) using machine learning models trained on historical accident records. The system: * Cleans and preprocesses real accident data * Handles severe class imbalance using SMOTE * Trains and evaluates multiple ML models * Allows users to interactively test predictions via a Streamlit web app 🗺️ Dataset * Source:Addis Ababa Road Traffic Accident Dataset * Country:Ethiopia 🇪🇹 * Target Variable:Accident_severity * `0` → Fatal * `1` → Serious * `2` → Slight Key Features Used Hour of the accident Number of vehicles involved Number of casualties Encoded accident characteristics (cause, road condition, etc.) 🧠 Models Trained The following models were trained and evaluated: | Model | Description | | ----------------------- | -------------------------- | | Logistic Regression | Baseline linear classifier | | Random Forest | Ensemble tree-based model | | Gradient Boosting | Best performing model | 📊 Evaluation Metric 🔍 Model Performance (Macro F1) | Model | Macro F1-score | | --------------------- | -------------- | | Logistic Regression | 0.31 | | Random Forest | 0.37 | | **Gradient Boosting** | **0.43** ⭐ | 🧪 Project Structure ``` ethiopia-road-accident-prediction/ │ ├── data/ │ ├── train_X.csv │ ├── train_y.csv │ ├── test_X.csv │ └── test_y.csv │ ├── src/ │ ├── preprocess.py │ ├── train.py │ └── evaluate.py │ ├── app.py ├── model_lr.pkl ├── model_rf_best.pkl ├── model_gb_ …