Rainfall Prediction Using Machine Learning And Time Series Forecasting Techniques

Abstract

In India, where 75% of agriculture business is reliant on precipitation as its principle wellspring of water, the time and measure of precipitation hold high significance and can influence the whole economy of the country. To forecast precipitation is one of the major studies in meteorological science. In this study, an attempt to study few statistical techniques and machine learning techniques for the prediction and estimation of meteorological parameters is made and based on that estimation, prediction of Precipitation is done. For the study, daily data set with meteorological parameters like maximum temperature, minimum temperature, relative humidity, wind speed, precipitation were taken into consideration. Validation was also performed to assess the accuracy of forecasting Model. Validation exercise shows the comparative usefulness of different techniques. The study shows that ARIMA model and Neural Network model works best for forecasting meteorological parameters like minimum temperature, maximum temperature and Random Forest model gives best classification accuracy compared to other machine learning algorithms for forecasting next monsoon season. Rainfall prediction is more complex and challenging using satellite image due to ever changing weather conditions, so an effort has also been made to predict the same.