AstraMeron/ethiopia-fi-forecast

# Ethiopia Financial Inclusion Forecasting (2026-2027) **Author:** Meron Tilahun (Mary) | 5th Year CSE, ASTU **Training:** 10 Academy KAIM (Batch 2026) ## 📌 Project Overview This project focuses on modeling and forecasting financial inclusion (FI) indicators in Ethiopia. It utilizes a **Unified Schema** to combine historical survey data with real-time market milestones (Telebirr, Safaricom, Fayda ID). --- ## 🛠 Task 1: Data Enrichment & Schema Mapping **Objective:** Enhance the raw dataset with recent market observations to improve forecast accuracy. - **Key Additions:** - Updated **Telebirr** usage data to EOY 2025 (58.6M users). - Integrated **Fayda Digital ID** mandatory policy as a key institutional event. - **Unified Schema:** All data was mapped to a standardized format involving `pillars`, `indicators`, and `impact_links`. ## 📊 Task 2: Exploratory Data Analysis (EDA) **Objective:** Analyze patterns and identify drivers of financial inclusion. - **Key Insights:** 1. **The +3pp Stagnation:** Account ownership grew only 3% (2021-2024) despite massive mobile money expansion, likely due to KYC barriers and dormancy. 2. **Infrastructure Lead:** 4G coverage growth consistently leads usage growth by 12–18 months. 3. **The ID Catalyst:** Analysis suggests the Digital ID rollout is the most significant non-market driver for 2026 growth. ## 🧪 Robustness & Reproducibility - **Error Handling:** Notebooks include `try-except` blocks for data loading and path validation. - **Data Validation:** Automated assertions check for duplicate record IDs and critical null values. - **Processed Output:** The final enriched dataset is available in `data/processed/ethiopia_fi_final_enriched.csv`. ## ⚠️ Limitations - **Data Sparsity:** Relying on 3-year Findex cycles creates temporal gaps. - **Proxy Use:** Operator data (registered accounts) may overrepresent unique users due to multi-SIM cards. ## ⚖️ Task 3: Event Impact Quantification **Objective:** Translate qualitative events ( …