Rainfall Forecasting Using Machine Learning Techniques: Case Study Surat, India

Natural risks are increasing due to an increase in the power, length, and frequency of extreme events. i.e., humans and the environment are under stress due to changes in the pattern of precipitation and temperature increases. Living circumstances are deteriorating due to a shortage of natural resources as population increase occurs in vulnerable regions of the world. In addition, Natural disaster is a factor to damage it. People migrating is another factor in it. The change in climate is fuel for the making this condition worse in developing country. The developed world has already dealt with these issues and made some effort to find solutions. The developed countries are facing financial loss because of higher usage of natural resources. So, there is need of forecasting model to overcome this uncertainty of weather. In this paper, we have collected 2,27,167 records of different areas of Surat city from year 2013 to 2016. We have predicted the rainfall of Surat, India. Our dataset has attributes like Relative Humidity (RH), Rainfall, Temp, Wind Dir and Wind Speed. We have preprocessed the data using techniques like data cleaning, data aggregation and data normalization. We have calculated Correlation Coefficient to gain those attributes which are having more correlation with rainfall. Then, we implemented Linear Regression and Neural Network model on preprocessed and un-processed data to predict rainfall in Surat city. We concluded that preprocessed data is having better Adjusted R2 than unprocessed data. We have generated result using matrices like RMSE, MSE, Adjusted R and R2.