samolubode/finhealth

Financial Health Prediction modelling for African MSMEs # MSME Financial Health Index (FHI) Prediction This project aims to predict the Financial Health Index (FHI) of small and medium-sized enterprises (SMEs) across Southern Africa (Eswatini, Lesotho, Malawi, and Zimbabwe). The model uses features such as business characteristics, owner demographics, financial behavior, and risk factors. ## Prerequisites - **Python 3.10+** - **Mac Users (Apple Silicon)**: Ensure `libomp` is installed for XGBoost. ```bash brew install libomp ``` ## Setup Instructions 1. **Create and Activate Virtual Environment**: ```bash python3 -m venv venv source venv/bin/activate ``` 2. **Install Dependencies**: ```bash pip install -r requirements.txt ``` ## Usage ### 1. Model Training To train the model, perform hyperparameter tuning, and evaluate performance using 5-fold Stratified Cross-Validation: ```bash python3 train.py ``` This script will: - Preprocess the data (handling missing values, feature engineering). - Run a Randomized Search for optimal XGBoost hyperparameters. - Save the final model to `model.pkl` and encoders to `encoders.pkl`. ### 2. Generate Predictions To generate predictions for the test set: ```bash python3 predict.py ``` This will create a `submission.csv` file in the project root. ### 3. Data Analysis To view feature importance and target distribution analysis: ```bash python3 analyze_imp.py ``` ## Project Structure - `preprocess.py`: Contains data cleaning and feature engineering logic. - `train.py`: Training pipeline with hyperparameter tuning. - `predict.py`: Inference script for generating test set predictions. - `analyze_imp.py`: Utility for analyzing model performance and feature importancia. - `data/`: Contains raw CSV data files (`Train.csv`, `Test.csv`, etc.).