The impact of climate change and food production on food products export earnings in Zimbabwe: a machine learning approach.

Abstract

Climate change is a threat to the agriculture-based economy in Zimbabwe due to its effect on rainfall, droughts, temperature changes, and natural disasters, which hinder production and export earnings. This research aims at exploring the effect of climate change and food production on food product export earnings in Zimbabwe using annual data for the period 1967–2025. Two machine learning models: extreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM) are used to estimate the relationship between climate factors, food production, and export earnings. Data was collected from ZIMSTAT, the Food and Agriculture Organization (FAO), and the World Bank while using lagged values and three-year rolling averages in the process of feature engineering to enhance accuracy. Results show that there is a very strong relationship between food production and export earnings (r = 0.68). Climate variability affects trade through its effect on food production, hence its effect on earnings from exports. Model predictions are highly accurate, XGBoost outperforms the LightGBM, giving an r² of 0.981 and a MAPE of 3.46%. Historical values of export earnings and food production are among the top features used in predicting future export performance.