MACHINE LEARNING-BASED WEATHER PREDICTION: A COMPARATIVE STUDY OF REGRESSION AND CLASSIFICATION ALGORITHMS

Abstract

Accurate weather forecasting is crucial across multiple industries like agriculture, transportation, and
disaster management, making it a key application for machine learning. This study explores the prediction of
various weather conditions—rain, sunshine, clouds, fog, drizzle, and snow—using a range of fundamental machine
learning techniques and boosting algorithms. Historical meteorological data, including temperature, humidity, wind
speed, and pressure, was used to train and evaluate these algorithms. Tests encompassed well-known methods such
as decision trees, random forests, naive Bayes, k-nearest neighbors, and support vector machines. Additionally,
boosting methods like XGBoost and AdaBoost were employed to improve forecast precision.
Results indicated that XGBoost and AdaBoost achieved the highest accuracies (87.86% and 87.33%, respectively)
compared to other methods tested. Validation through ROC Curve Analysis and Lift Curve Analysis demonstrated
superior performance of XGBoost and AdaBoost in terms of true positive rate, false positive rate, and lift.[1,2]