yevqlzw/African_Crisis_Prediction

An end-to-end machine learning pipeline for early warning of banking, currency, and sovereign debt crises across 13 African nations - one year in advance. # 🌍 African Systemic Crisis Prediction > An end-to-end machine learning pipeline for early warning of banking, currency, and sovereign debt crises across 13 African nations - one year in advance. --- ## Overview Predicting economic crises is a core challenge in risk management, policy-making, and international finance. This project builds a supervised binary classifier that forecasts the onset of a **systemic crisis** one year ahead, replicating the analytical workflow of a risk analyst at a central bank or international organisation. The dataset spans **13 African countries** from the late 19th century to 2014, providing a rich historical panel for time-series classification under severe class imbalance. > **Why this matters:** Early Warning Systems (EWS) for financial crises are actively used by the IMF, World Bank, and central banks globally. This project demonstrates how historical macroeconomic data can be transformed into actionable risk signals — without any future data leakage. --- ## Dataset **Source:** `african_crises.csv` - a curated panel dataset of African economic history. | Attribute | Detail | |-----------|--------| | Countries | Algeria, Angola, Central African Republic, Ivory Coast, Egypt, Kenya, Mauritius, Morocco, Nigeria, South Africa, Tunisia, Zambia, Zimbabwe | | Period | ~1860 – 2014 | | Observations | 1 059 country-year rows | | Target rate | ~7.7 % systemic crisis years (severe class imbalance) | **Key raw columns:** | Column | Description | |--------|---------------------------------------------------------| | `country`, `year` | Panel identifiers | | `exch_usd` | Exchange rate (local currency per USD) | | `inflation_annual_cpi` | Annual CPI inflation rate (%) | | `domestic_debt_in_default` | Domestic debt default flag (0/1) | | `sovereign_external_debt_default` | Sovereign externa …