Faysalseifu/End-to-End-Insurance-Risk-Analytics-Predictive-Modeling

Analyze South African car insurance data to identify low-risk segments. Perform EDA, A/B hypothesis testing (province, postal code, gender), build claim severity & premium prediction models using XGBoost, and implement DVC + Git workflows. Deliver actionable insights to optimize pricing and marketing # End-to-End-Insurance-Risk-Analytics-Predictive-Modeling Analyze South African car insurance data to identify low-risk segments. Perform EDA, A/B hypothesis testing (province, postal code, gender), build claim severity & premium prediction models using XGBoost, and implement DVC + Git workflows. Deliver actionable insights to optimize pricing and marketing. ## Quick start ```bash python -m venv .venv source .venv/bin/activate # Windows: .venv\Scripts\activate pip install -r requirements.txt ``` ## DVC workflow (local remote) - Install: `pip install dvc` - Initialize: `dvc init` - Create local remote directory: `mkdir -p ../.dvc-storage` (from repo root) - Configure default remote: `dvc remote add -d localstorage ../.dvc-storage` - Track data: `dvc add data/insurance.csv data/raw_data.txt data/processed_data.txt` - Push artifacts: `dvc push` - Commit to git: add `.dvc` files plus `data/.gitignore`; never commit raw data files. ## Running EDA notebook - Open `notebooks/Exploratory Data Analysis.ipynb` in Jupyter Lab or VS Code. - Run cells sequentially; plots cover distributions, outliers, smoker impact on charges, and feature correlations. ## CI/CD - GitHub Actions workflow (`.github/workflows/ci.yml`) installs dependencies, compiles Python files, and runs pytest (tolerant if no tests are present). Triggers on pushes/PRs to `main` and `task-1`. ## Data notes - Current sample dataset (`data/insurance.csv`) includes demographics, BMI, smoker flag, region, and charges. It lacks premium, claim frequency, vehicle, geography granularity, and dates—so loss-ratio and temporal trend analyses from the brief cannot be computed until richer data is added.