Rainfall prediction by using machine learning approach

This study utilized machine learning to forecast rainfall levels and identify influential atmospheric elements. Python software was employed to analyze historical rainfall data, focusing on 16 selected attributes. The study aimed to predict rainfall amounts and determine the features that influence rainfall the most. Its significance lies in optimizing agricultural practices, enhancing preparedness for heavy rains, and supporting infrastructure planning, sustainable agriculture, and economic resilience. MLR algorithms were utilized and evaluated using the dataset. The data underwent preprocessing to handle missing values and create an organized dataset suitable for model building. The rainfall dataset was divided into a training set (70%) and a testing set (30%). The resulting model's predictive performance for rainfall was assessed by calculating the R-squared value. RFE was employed to select the most relevant two features, ranking them based on relevance and eliminating less important ones. The model developed using RFE demonstrated a reasonably good fit to the testing data, achieving an R-squared value of0.1316. The RMSE indicated an average difference of around 0.11385 units between predicted and actual values. The errors between predicted and actual values were reasonably minimal, with a MSE of0.01919 and a MAE of0.07436. The model accounted for approximately 13% of the variation in rainfall. To conclude, the general aims of this study, where the factors that influence most rainfall have been found, by comparing the model and the Recursive Features Selection (RFE) process. The specific objectives of this study were fully addressed as the MSE and MAE value is low.