The occurrence of rainfall is an outcome of various natural factors such as temperature, humidity, cloudiness, wind speed, etc. Rainfall prediction is a major concern for meteorological department as it is closely associated with the economy and sustenance of human life. In this work, we use regression techniques and statistical modelling to predict the rainfall intensity of Coonoor in Nilgiris district, Tamil Nadu. It is a comparative study of various regression techniques based on the Relative error. The regression techniques used for prediction are Support Vector Regression (SVR), Random forest (RF) and Decision Tree (DT). The parameters considered for training the model includes the daily recorded temperature, humidity, cloud speed, wind speed and wind direction of Coonoor. The rainfall prediction model was made more efficient by including the rainfall intensities of nearby stations within an area of 7 km 2 , The developed forecasting models were analysed on the basis of R-square and Adjusted R-square values. A statistical model was developed out of all the techniques by generating the regression equation used for prediction of rainfall by each of the model. The proposed models were implemented in Python platform. The prediction model developed by the RF regression technique was found out to be a better and efficient model compared to SVR and DT models.
|Journal||Data powered by Typeset2018 Second International Conference on Inventive Communication and Computational Technologies (ICICCT)|
|Publisher||Data powered by TypesetIEEE|