Evaluating the Performance of Support Vector Machines for Weather Forecasting

Abstract

Accurate weather forecasting is crucial for sectors such as agriculture, disaster management, and transportation due to the nonlinear and complex nature of weather systems that traditional meteorological methods often struggle to capture. The emergence of machine learning has opened up new possibilities for improving the accuracy of weather predictions. Objective of the research is trying to assess the efficacy of Support Vector Machines (SVMs) in forecasting weather conditions by utilizing a reliable Kaggle dataset. The SVM model integrates a variety of kernel functions, encompassing lots of functions. Its effectiveness is evaluated in comparison to alternative ML models like neural networks and decision trees. The proposed methodology includes essential data preprocessing steps like normalization, dimensionality reduction through Principal Component Analysis (PCA), and meticulous model selection. The findings indicate that the SVM utilizing an RBF kernel exhibits superior performance compared to alternative models, the accuracy, recall, and F1-score. This study showcases the potential of SVMs as a viable tool for weather forecasting, contributing to advancements in predictive meteorology. Future research could explore automated hyperparameter tuning techniques and real-time weather predictions to further enhance model performance.