Machine Learning-based Weather Prediction: A Comparative Study of Regression and Classification Algorithms

Accurate weather forecasting is essential in many industries, including agriculture, transportation, and disaster management, making it a prime use case for machine learning algorithms. In this study, we investigate how to forecast several types of weather, including rain, sunshine, clouds, fog, drizzle, and snow, using a variety of fundamental machine learning methods and boosting algorithms. To train and evaluate the various algorithms, we utilized a dataset made up of historical meteorological data, including characteristics like temperature, humidity, wind speed, and pressure. We performed tests on many machine learning methods, some of which you may be familiar with: decision trees, random forests, naive bayes, k-nearest neighbors, and support vector machines. We also used boosting techniques like XGBoost and AdaBoost to further enhance the precision of our forecasts. Our results indicated that XGBoost and AdaBoost, two popular boosting algorithms, achieved the highest levels of accuracy (87.86% and 87.33%) compared to the other algorithms we tested. The findings were verified using ROC Curve Analysis and Lift Curve Analysis, which demonstrated that the XGBoost and AdaBoost models performed better in terms of true positive rate, false positive rate, and lift.