KholoTheCoder/South-Africa-Inflation-Forecasting-ML

Machine learning project that forecasts SA inflation using CPI time-series data. # South-Africa-Inflation-Forecasting-ML Machine learning project that forecasts SA inflation using CPI time-series data. OVERVIEW This project builds a complete Inflation Forecasting System using South Africa's historical Consumer Price Index (CPI) values. Using machine learning, time-series modeling, and data visualization, it predicts future inflation trends — a skill crucial for: SARB (South African Reserve Bank) Financial analysts Economists Data scientists in finance SARS data analytics roles The model uses CPI data from 1960–2023 sourced from the World Bank API. | Area | Tools Used | | ------------------- | -------------------------------------------------- | | Programming | Python, Google Colab | | ML Models | ARIMA, Auto-ARIMA | | Data | World Bank CPI (1960–2023) | | Libraries | pandas, matplotlib, seaborn, statsmodels, pmdarima | | Evaluation | MAE, RMSE | | Visualization | Seasonal Decomposition, Line charts | | Deployment (future) | Streamlit / API-ready | FEATURES DATA ENGINEEERING Automated CPI downloading via stable World Bank API Complete cleaning + formatting Time-series indexing Missing value handling EXPLORATORY DATA ANALYSYS (EDA) CPI historical trend visualization Seasonal decomposition Trend + seasonality extraction MACHINE LEARNING Auto-ARIMA model selection Train/Test split Automated forecasting Residual diagnostics EVALUATION Mean Absolute Error (MAE) Root Mean Squared Error (RMSE) Forecast vs actual plot VISUAL OUTPUTS. The notebook generates: 1. CPI Trend (1960–2023) Shows inflation growth over decades. 2. Seasonal Decomposition 4-panel view: Observed Trend Seasonal Residual 3. Forecast Plot Training data Test (a …