idreamofunicorns/FinAccess-Predictor

A machine learning project that predicts bank account ownership in East Africa using demographic and socio-economic features. This project applies feature selection, SMOTE resampling, and hyperparameter tuning to improve model performance, helping financial institutions and policymakers identify unbanked populations and promote financial inclusion. # Predicting Bank Account Ownership in East Africa *Using Machine Learning to Enhance Financial Inclusion ## Project Overview Financial inclusion is critical for economic development, yet many individuals in East Africa remain unbanked. This project leverages machine learning to predict bank account ownership using demographic and socio-economic data. Insights from this model can help financial institutions, policymakers, and international organizations identify key factors influencing financial inclusion. ## Business Understanding Financial inclusion promotes economic growth and poverty reduction. The World Bank recognizes it as a key driver for achieving 7 out of 17 Sustainable Development Goals (SDGs). This project aims to develop a predictive model to classify individuals as either banked (1) or unbanked (0). ### Business Objective **Goal**: Predict whether an individual owns a bank account. **Use Case**: Assist financial institutions and policymakers in targeting the unbanked population. ### Key Business Questions 1.Who does not have a bank account? 2.What factors influence bank account ownership? 3.How can the model best support financial inclusion initiatives? ### Success Criteria 1.Model Accuracy: Ensure high predictive performance. 2.Actionable Insights: Identify key features influencing bank account ownership. 3.Business Impact: Support financial institutions in targeting unbanked individuals. ### Dataset Overview Total Records: 23,524 individuals from Kenya, Uganda, Tanzania, and Rwanda. Target Variable: Bank account ownership (1 = Yes, 0 = No). Features: Categorical: Country, location type, relationship with head, marital status, education level, job type. Numerical: Age, household size, year. Binary: Gender, cellphone access. ### Data Preprocessing Handled class imbalance using SMOTE (Synthetic Minority Over-sampling Technique). Feature selection to retain the most relevant predictors. Applied data cleaning & encoding for categorical variables. # …