A Comparative Study of Regressors and Stacked Ensemble Model for Daily Temperature Forecasting: A Case Study of Senegal

Abstract

Over the Sahel region, air temperature is anticipated to rise by 2.0 to 4.3 ∘C by 2080. This increase is likely to affect human life. Thus, air temperature forecasting is an important research topic. This study compares the performance of stacked Ensemble Model and three regressors: Gradient Boosting, CatBoost and Light Gradient Boosting Machine for daily Maximum Temperature and Minimum Temperature forecasting based on the five lagged values. Results obtained demonstrate that the Ensemble Model outperformed the regressors as follows for each parameter; Maximum Temperature: MSE 2.8038, RMSE 1.6591 and R2 0.8205. For Minimum Temperature: MSE 1.1329, RMSE 1.0515 and R2 0.9018. Considering these results, Ensemble Model is observed to be feasible for daily Maximum and Minimum Temperature forecasting.