Abstract
Every farmer requires access to rainfall prediction (RP) to continue their exploration of harvest yield. The proper use of water assets, the successful collection of water, and the successful pre-growth of water construction all depend on an accurate assessment of rainfall. The prediction of heavy rain and the provision of information regarding natural catastrophes are two of the most challenging factors in this regard. In the twentieth century, RP was the most methodically and technically complicated issue worldwide. Weather prediction may be used to calculate and analyse the behaviour of weather with unique features and to determine rainfall patterns at an exact locale. To this end, a variety of methodologies have been used to determine the rainfall intensity in Saudi Arabia. The classification methods of data mining (DM) approaches that estimate rainfall both numerically and categorically can be used to achieve RP. This study, which used DM approaches, achieved greater accuracy in RP than conventional statistical methods. This study was conducted to test the efficacy of several machine learning (ML) approaches for forecasting rainfall, utilising southern Saudi Arabia's historical weather data obtained from the live database that comprises various meteorological data variables. Accurate crop yield predictions are crucial and would undoubtedly assist farmers. While engineers have developed analysis systems whose performance relies on several connected factors, these methods are seldom used despite their potential for precise crop yield forecasts. For this reason, agricultural forecasting should make use of these methods. The impact of drought on crop yield can be difficult to forecast and there is a need for careful preparation regarding crop choice, planting window, harvest motive, and storage space. In this study, the relevant characteristics required to predict precipitation were identified and the ML approach utilised is an innovative classification method that can be used determine whether the predicted rainfall will be regular or heavy. The outcomes of several different methodologies, including accuracy, error, recall, F-measure, RMSE, and MAE, are used to evaluate the performance metrics. Based on this evaluation, it is determined that DT provides the highest level of accuracy. The accuracy of the Function Fitting Artificial Neural Network classifier (FFANN) is 96.1%, which is higher than that of any of the other classifiers currently used in the rainfall database.