Prediction of Rainfall Using Different Machine Learning Regression Models

Abstract

Rainfall is one of the major climatic aspects of India. It has both pros and cons. Rainfall prediction is a regression problem, and if we predict the amount of rainfall in advance, we can assess the impact of it and decrease the risk of loss. One of the major problems today to predict rainfall using machine learning is accuracy. To increase that accuracy, we can first use data preprocessing techniques and select the best attributes and then use all well-known ML models and select the accurate model from the existing one. We propose the process to select model for predicting rainfall using machine learning techniques in India. Here, we will perform the visual examination of distinct data variables to uncover all the patterns using data exploration techniques. To analyze rainfall trends, scatter plots, histograms, bar charts, box plots, and heatmaps are used. Multiple linear regression, support vector regression, random forest regression, lasso regression, and ridge regression are five regression models that can be used to model data. Finally, the selected model predicts rainfall with highest accuracy. Different machine learning regression models for prediction were developed, and after comparative analysis, based on performance evaluation metrics, a best model for predicting rainfall for a subdivision is implemented. Also, the rainfall trends in India can be studied using exploratory data analysis and data visualization.