bcudjoe/mobile-money-ghana-forecasting

# Predicting Mobile Money Transactions and Digital Financial Inclusion Growth in Ghana Forecasting Ghana's mobile money growth and explaining its drivers with machine learning and publicly available financial-access indicators. **Author:** Benjamin Asomia Cudjoe · **Course:** QM640 Data Analytics Capstone, Walsh College · **Term:** Third Term 2026 --- ## Overview This project studies digital financial inclusion in Ghana along two tracks that are kept separate because they use different units of analysis: - **National time-series track** — a monthly Bank of Ghana payment panel of **84 months (January 2019 – December 2025)** is used to forecast mobile money transaction value and active-account growth six to twelve months ahead. - **Individual cross-sectional track** — **2,000 pooled Global Findex respondents** (1,000 from the 2021 wave and 1,000 from the 2025 wave) are used to identify which socioeconomic and financial factors drive individual adoption. The two tracks are analysed on their own and their results are combined only at the interpretation stage to produce policy recommendations. ## Research questions | RQ | Question | Track | Method | |----|----------|-------|--------| | RQ1 | Which factors most strongly drive mobile money adoption? | Individual (Findex, n = 2,000) | Logistic regression + tree model with driver ranking | | RQ2 | Can models forecast transaction value and account growth? | National monthly (n = 84) | Lag-feature regression, with SARIMA/Prophet and gradient boosting to follow | | RQ3 | Which approach forecasts most accurately? | National monthly | Holdout comparison (RMSE/MAE/MAPE/R²) | | RQ4 | What policy actions improve inclusion? | Synthesis | Driver-based scenario analysis | ## Repository structure ``` mobile-money-ghana-forecasting/ ├── data/ │ ├── raw/ # source downloads │ │ ├── monthly_series.xlsx # BoG monthly panel (raw columns) │ │ ├── findex_ghana_2021.xlsx # Findex 2021 w …