Leveraging Machine Learning for Accurate Prediction of Extreme Temperature Events

Abstract

Heatwaves and cold waves are periods of overly hot or cold climate conditions that are becoming more frequent due to climate change with negative impacts, including threats to human life, crops, and human science facilities. These events are challenging to predict with conventional weather models; therefore, we are turning to ML approaches. This study examines and contrasts several machine learning algorithms for forecasting extreme temperatures, including gradient boosting, random forest, elastic net, linear regression, SVR, XGBoost, and CatBoost. According to these metrics, we assessed the performance of the models mentioned above with the help of historical weather data in terms of root mean square error (RMSE) and mean absolute error (MAE). The CatBoost Regressor is shown to be the best in terms of the best balance between accuracy scores and the minimal preprocessing required. The outcomes presented in the paper indicate that CatBoost and other ML algorithms can improve extreme temperature prediction, which is useful for generating early alerts and developing climate risk mitigation strategies. The evaluation of the suggested model looked into the mean square error (MSE), root mean square error (RMSE), and minimum absolute error (MAE) that the Catboost regressor achieved. The results showed that the determination coefficient, R2, was MSE = 0.6725, MAE = 0.6047, correlation = 0.9631, and P-value of correlation = 0.0000, respectively.