Daksam/commodity-export-prediction

This is the repository for my export prediction project. Topic: Predicting monthly export commodity volumes for Ghana using tree models and sequence models # Predicting monthly export commodity volumes for Ghana using tree models and sequence models ## This is the repository for my project. Predicting monthly export commodity volumes is a critical challenge for resource-dependent economies such as Ghana, where cocoa, gold, and petroleum collectively account for the majority of foreign exchange earnings and fiscal revenue. This study develops and evaluates a comparative forecasting framework that applies four models — Prophet, Random Forest, XGBoost, and an Encoder-Decoder Long Short-Term Memory (LSTM) network to monthly export volume data for all three commodities over the period 2010 to 2025. The data was sourced from the Ghana Statistical Service and UN Comtrade and augmented with World Bank commodity prices (Pink Sheet) and Bank of Ghana exchange rate data. Models are evaluated using RMSE and MAPE at forecast horizons of 1, 3, 6, and 12 months ahead under a strict walk-forward temporal validation protocol. The results indicate that XGBoost outperformed the other three models with MAPE = 3.67% for cocoa and MAPE = 3.28% for gold. In the case of petroleum, LSTM is the best performing model with MAPE = 2.89%.