HELI: An Ensemble Forecasting Approach for Temperature Prediction in the Context of Climate Change

Abstract

Climate change is a critical challenge, demanding the development of effective methods for temperature forecasting. Statistical and machine learning models emerge as promising alternatives. However, there is no widely accepted superior method; ensemble approaches integrate strategies that take advantage of each forecasting method. Ensemble methodologies combine methods, weighing their participation to integrate each of them. Forecasting researchers have shown that evolutionary algorithms are highly effective in achieving an ensemble that is at least as effective as the best single method. This paper presents HELI, a forecasting methodology designed to forecast the climate temperature variable; its architecture is modular, aiming to provide a flexible forecasting application in the climate change area. We present experimentation and a hypothesis test for a region in Mexico City and show HELI's competitiveness compared to leading strategies. Besides, we present experiments with other climate change variables that show HELI flexibility in the context of climate change.