FORCASTING COFFEE PRICE USING STACKING ENSEMBLE APPROACH IN CASE OF ETHIOPIAN COMMODITY EXCHANGE

Coffee is one of the most widely traded commodities in the world, with prices that can greatly impact the economies of coffee-producing countries such as Ethiopia. The Ethiopian Commodity Exchange (ECX) is a key player in the coffee market in southwestern Ethiopia, and accurate forecasting of coffee prices is critical for farmers, traders, and other stakeholders. This research addresses the challenge of accurately forecasting coffee prices in the context of the ECX. The study investigates the effectiveness of a stacking ensemble approach in combining the predictions of multiple models to make the final forecast. The research findings show that coffee plants in the region are typically grown by smallholder farmers at a density of around 30,000 plants per hectare, with the potential to yield up to millions of kilograms of coffee per year, including both specialty and commercial grades. The performance of individual models was evaluated, with the Random Forest algorithm showing the strongest predictive power, achieving a Mean Squared Error (MSE) of 0.06 and an R-squared value of 0.99. In contrast, the Linear Regression model struggled, with an MSE of 7.48 and an R-squared of -0.01. The Gradient Boosting model also performed well, with an MSE of 0.77 and an R-squared of 0.90. However, the true strength of the research lies in the stacking ensemble approach, which was able to optimally combine the predictions of the individual base learners. The Stacking Regressor meta-model delivered the best overall performance, with an MSE of 0.11 and an Rsquared of 0.98, significantly outperforming the standalone models. These findings highlight the power of ensemble learning techniques, such as stacking, in leveraging the complementary strengths of diverse modeling approaches to achieve superior forecasting accuracy, which is crucial in the coffee sector where accurate price predictions can have a substantial impact on the livelihoods of smallholder farmers and other stakeholders.