<p>Due to the increasing effects of climate change, Morocco is experiencing major disruptions in its rainfall pattern, threatening water resource management, agricultural planning and environmental sustainability. This growing instability makes accurate prediction of rainfall an essential requirement. However, traditional prediction models are often inefficient and computationally expensive, particularly when applied to the actual heterogeneous and large-scale climate datasets. In response to these limitations, our study investigates the use of machine learning models as a viable and efficient alternative. We introduce a structured taxonomy that classifies machine learning algorithms into two main categories: parametric and non-parametric, offering a new view on model behavior in climate prediction. Prior to modelling, the dataset undergoes a comprehensive preparation process including data cleaning and feature selection, with the aim of optimizing the performance of each algorithm. A comparative evaluation is then carried out using Root Mean Squared Error and Mean Absolute Error as performance indices. Experimental results show that non-parametric models consistently outperform parametric models, owing to their ability to reveal complex and non-linear relationships between meteorological features.</p>