stephenkinuthia-cell/Zindi-Financial-Inclusion-in-Africa

A machine learning project for predicting bank account ownership using demographic data, with a full pipeline covering EDA, feature engineering, model training, evaluation, and explainability. # Financial Inclusion in Africa – Zindi Challenge ## Project Overview This project is an **end-to-end machine learning pipeline** developed for the Zindi learning challenge **“Financial Inclusion in Africa”** . The goal is to predict whether an individual is likely to have or use a bank account based on demographic and socio-economic characteristics across four East African countries: **Kenya, Rwanda, Tanzania, and Uganda** . The project goes beyond model training to include **structured preprocessing, model comparison, automatic model selection, interpretability using SHAP, and preparation for deployment via a dashboard** . --- ## Problem Statement Financial inclusion is a key driver of economic growth and human development. However, a large proportion of adults in East Africa do not have access to formal banking services. **Objective:** Predict whether an individual has a bank account (`Yes = 1`, `No = 0`) using survey data, and identify the key factors driving financial inclusion. --- ## Machine Learning Framing * **Task type:** Binary classification * **Target variable:** `bank_account` * **Evaluation focus:** Recall (due to class imbalance) * **Metric used by competition:** Mean Absolute Error (MAE) Because the dataset is imbalanced (fewer people have bank accounts), **recall is prioritized** to avoid missing individuals who are banked. --- ## Project Structure ```text financial-inclusion-zindi/ │ ├── data/ │ ├── raw/ # Original Zindi datasets (ignored in Git) │ │ ├── Train.csv │ │ └── Test.csv │ ├── notebooks/ │ ├── 01_eda.ipynb # Exploratory Data Analysis │ ├── 02_feature_engineering.ipynb │ └── 03_modeling.ipynb # Model training, evaluation & SHAP │ ├── src/ │ ├── preprocess.py # Feature engineering pipelines │ ├── train.py # Model training & auto-selection │ └── predict.py # Inference & submission generation │ ├── models/ │ ├── __init__.py │ ├── logistic.py │ ├── …