Development of machine learning models for predicting average annual temperatures

Abstract

This study assesses machine learning models for predicting Antarctica's average annual temperatures, addressing the challenge of accuracy in remote and variable climatic conditions. Four models were compared: linear regression, random forest regressor, decision tree regressor, and gradient boosting, utilizing data from diverse Antarctic stations. Results indicate the superiority of specific models tailored to individual stations, with the random forest model demonstrating exceptional performance across most metrics. This emphasizes the significance of geographical specificity in improving climate prediction accuracy. The research underscores machine learning's potential in climate change forecasting, advocating for tailored approaches in environmental modeling.