A support vector regression model for forecasting rainfall

Rainfall prediction is a very important part of weather forecasting. In countries like Bangladesh; which has several seasons a year, rainfall prediction is really a key factor for many sectors. Rainfall data is a time series data and it changes time to time as climate and season changes. Moreover, rainfall depends on several factors as flow of wind, humidity etc., it is very challenging to make a hundred percent perfect prediction. This paper exhibits a robust rainfall prediction technique in view of the recent rainfall data of Bangladesh utilizing Support Vector Regression (SVR), a relapse methodology of Support Vector Machine (SVR). The collected raw data wasn't prepared for using as input of algorithm, thus it had been pre processed manually to suit into the algorithm, then fed to the algorithm. The evaluation results of the study conducted on the data shows that the projected technique performs higher than the conventional frameworks in term of accuracy and process running time. The proposed approach yielded the utmost prediction of 99.92%.