Leveraging Meteorological Data and Machine Learning for Improved Rainfall Forecasting in Australia
Akash Das, K. M. Asieb Hasan · 2024
This study explores the application of machine learning techniques to predict rainfall in Australia, addressing the challenges posed by the country’s highly variable climate. Utilizing a comprehensive dataset spanning approximately 10 years of daily weather observations from multiple Australian locations, we implemented and compared six machine learning models: Random Forest, XGBoost, Logistic Regression, K-Nearest Neighbors (KNN), Artificial Neural Network (ANN) Classifier, and Naive Bayes. The models were evaluated on their ability to predict the binary outcome of rainfall occurrence for the following day. Our results demonstrate that ensemble methods, specifically Random Forest and XGBoost, achieved the highest accuracy at 86%, closely followed by Logistic Regression at 85%. All models showed stronger performance in predicting non-rainy days compared to rainy days, reflecting the dataset’s inherent class imbalance. The ANN Classifier exhibited the highest recall for rain prediction, suggesting its potential utility in identifying possible rainfall events.