Rainfall Prediction Using Exploratory Data Analysis and Machine Learning
Abstract
Forecasting rainfall is essential for managing water resources, agriculture, and preventing disasters. In this study, we suggest a machine learning-based rainfall prediction system that improves forecast accuracy by addressing a number of real-world data issues. In order to identify and handle missing values via mode imputation, data pretreatment techniques were first used. Using statistical methods (mean ± 3*standard deviation), outliers were found and dealt with properly to enhance model performance. Using the SMOTE (Synthetic Minority Oversampling Technique) framework, the dataset’s class imbalance was addressed. To facilitate model selection and comprehend inter-feature interactions, a correlation matrix was created. We have evaluated three different machine learning algorithms, which are Random Forest, Logistic Regression, and Decision Tree. The highest accuracy, 93%, has been achieved by Random forest, whereas the accuracies of LR and DT are 87% and 85%, respectively.