GIVEN-CHINYAMA/zambia-inflation-forecasting-app

Interactive Streamlit dashboard for forecasting Zambia's inflation using ARIMAX, XGBoost, Prophet, and LSTM models. Features real-time economic data integration, model benchmarking, interactive visualizations, and multi-year inflation forecasts. # 📊 Zambia Inflation Forecasting App An interactive **Streamlit dashboard** for forecasting Zambia's annual inflation rate using a hybrid ensemble of four machine learning and statistical models — **ARIMAX**, **XGBoost**, **Facebook Prophet**, and **LSTM**. The app integrates real-time macroeconomic data from the World Bank API and Yahoo Finance, enabling data-driven inflation analysis and multi-year forward forecasting. --- ## 🗂️ Table of Contents - Overview - Features - Models - Data Sources - Project Structure - Installation - Usage - Model Results - Key Insights - Author --- ## 🔍 Overview Zambia's inflation rate is influenced by a complex interplay of macroeconomic variables — particularly the **USD/ZMW exchange rate** and **copper commodity prices**, given the country's reliance on copper exports. This project builds a rigorous forecasting pipeline that: 1. Pulls live economic data from the **World Bank API** and **Yahoo Finance** 2. Performs **Exploratory Data Analysis (EDA)**, stationarity testing, and feature engineering 3. Trains and evaluates four distinct forecasting models 4. Benchmarks all models on RMSE, MAE, and MAPE 5. Generates a **3-year forward inflation forecast (2025–2027)** using the best-performing model 6. Presents all results through an **interactive Streamlit dashboard** --- ## ✨ Features - 📡 **Real-time data integration** — live pulls from World Bank API and Yahoo Finance - 🤖 **Four forecasting models** — statistical, machine learning, and deep learning approaches - 📈 **Interactive visualizations** — time series plots, confidence intervals, model overlays - 🏆 **Automated model benchmarking** — ranked comparison across RMSE, MAE, and MAPE - 🔭 **Multi-year forward forecasting** — 3-year inflation outlook with 95% confidence bands - 🧪 **Stationarity testing** — Augmented Dickey-Fuller (ADF) tests with automatic differencing - 📊 **Feature importance analysis** — XGBoost feature contribution breakdown --- ## 🤖 Models | Model …