Evaluation of the Performance of Machine Learning and Deep Learning Techniques for Predicting Rainfall: An Illustrative Case Study from Australia

Rainfall is a major factor in our ecological and environmental balance for a variety of reasons, including economy, agriculture, and cleanliness. It supplies the planet with essential fresh water, especially in areas where groundwater resources are scarce. Hence, a dependable prediction model for rainfall is essential, as it can help predict flooding and monitor pollutant levels. Historically, weather predictions were made using meteorological satellites. But now, with advancements in technology and data analysis, machine learning has been utilized in weather forecasting. However, accurately predicting rainfall remains a complex task and existing methods depend on complex models that may incur high costs due to their extensive computational requirements. This research assesses the effectiveness of both conventional machine learning algorithms and deep learning techniques as potential options, by conducting a comprehensive comparison using a uniform case study that analyzed ten years of rainfall data collected from various regions in Australia. Through the comparisons and evaluations, we aim at finding the most feasible method for the detection of weather patterns. The models' performance is measured using metrics such as loss, Mean Absolute Error, Mean Squared Error and Mean Squared Logarithmic Error. The results show that the proposed CNN model is the most accurate among all the models.