Solih06/ethiopia-fi-forecast

An event-augmented macro-forecasting system and interactive Streamlit dashboard designed to project adult account ownership trajectories in Ethiopia through 2027/2030, evaluate policy interventions (e.g., Fayda Digital ID, Rural Agent Banking expansion), and break down predictions using game-theoretic SHAP model explainability. # 🇪🇹 Ethiopia Financial Inclusion Forecast & Model Explainability Engine An event-augmented macro-forecasting system and interactive Streamlit dashboard designed to project adult account ownership trajectories in Ethiopia through 2027/2030, evaluate policy interventions (e.g., Fayda Digital ID, Rural Agent Banking expansion), and break down predictions using game-theoretic SHAP model explainability. --- ## Business Problem Despite rapid expansion in telecommunications and mobile money platforms (with registered mobile wallets exceeding 65 million), unique adult account ownership in Ethiopia remains fragmented at **49.0% (2024)**. Policy makers and financial consortiums face key operational hurdles: * **Target Gap Uncertainty:** Disconnect between raw digital account registrations and actual unique adult financial inclusion targets (60.0% by 2027 under NFIS-II). * **Policy Shock Blind Spots:** Lack of real-time tools to simulate how macroeconomic headwinds (inflation) or digital infrastructure rollouts (Fayda ID) shift long-term inclusion trajectories. * **Black-Box Decision Making:** Lack of explainable AI diagnostics to verify whether predictive models are driven by genuine structural levers or perpetuating demographic proxy bias. --- ## Solution Overview This system bridges sparse demand-side survey data (World Bank Global Findex) and high-frequency supply-side administrative indicators (National Bank of Ethiopia, EthSwitch) through an end-to-end framework: 1. **Unified Relational Data Framework:** Standardizes historical World Bank Findex survey series (2011–2024) across `observation`, `event`, `impact_link`, and `target` schemas. 2. **Event-Augmented Forecaster (`FIForecaster`):** Parametric curve fitting combined with policy impact additive adjustments to project Base, Optimistic, and Pessimistic scenarios. 3. **SHAP Model Explainability Engine:** Deconstructs global feature importance, local prediction waterfall breakdowns, and demographic bias dia …