Comparing Predictive Capabilities of Machine Learning Models in Weather Forecasting

The unpredictable and dynamic nature of weather poses great challenges in various fields of human activity such as agriculture, transportation, and outdoor activity planning. Therefore, the ability to accurately predict the weather is very important. Despite the availability of weather prediction systems, conventional approaches that rely on numerical and statistical methods often have limitations. These limitations include lack of flexibility in dealing with imbalanced data, slow response to data updates, as well as limitations in processing interacting non-linear features. Machine learning (ML) has become a versatile alternative that offers adaptability for improving weather forecast accuracy. This study seeks to assess the classification performance of several machine learning algorithms, specifically Decision Tree, Random Forest, Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost), in analyzing weather related parameters. The performance evaluation is conducted using performances metrics such as accuracy, precision, recall, and F1-Score. Results indicated that the XGBoost model consistently outperformed other models with 78.01% accuracy, 80.15% precision, 78.01% recall, and 78.90% F1-score. XGBoost proved to be able to handle data complexity, capture non-linear patterns, and cope better with class imbalance. This research shows that machine learning approaches, particularly XGBoost, provide great potential to replace or complement traditional, less adaptive weather prediction systems. With proper integration, ML-based weather prediction systems can become more responsive, accurate, and reliable in facing current and future weather prediction challenges.