Rainfall prediction using support vector regression in Udupi region Karnataka, India

Abstract

The hydromatereological processes are examined through analysis of temporal rainfall variability. India is an agricultural land and its economy is mainly dependent on timely rains to produce good harvest. The amount of rainfall varies with regional and temporal variation in distribution. The present research has been conducted to predict the temporal variations in rainfall in Udupi district, Karnataka, India using support vector regression (SVR) model and to validate the findings using actual rainfall records. The data has been collected from the statistical department, Udupi district, Government of Karnataka, India. The prediction accuracy of SVR based rainfall prediction model depends on tuning of algorithmic-based parameters. The parameter optimization is performed using grid search to select the optimal values of hyperparameters. The analysis was performed for the year 2018 based on the training dataset from 2000-2017. It is observed that there is a decreasing trend in total annual rainfall in 2018 and it is concluded that the average yearly rainfall has declined during the years 2018 and 2019. The rainfall predicted results were validated with actual records. The SVR based rainfall prediction model will predicts the rainfall accurately for application in agricultural sector.