Long term prediction of rainfall in Andhra Pradesh with Deep learning

Debnath Bhattacharyya*, Kalam Swathi, N. Thirupathi Rao, Nakka Marline Joys Kumari

Department of computer science and engineering, Koneru Laksmaiah Education Foundation, Greenfield, Vaddeswaram, Guntur, Andhra Pradesh, India

ABSTRACT

Rainfall is the major concern for almost types of people in the society. It helps different types of people in the society in different means. For some people, it’s the source of providing drinking water. For others like farmers, it’s the source of their livelihood and for all other human beings too. As farmers grow the farm and rice will be produced for the human being to eat. As a whole, the rainfall plays a key role for almost all kinds of people living in the society. Prediction of this rainfall is always an interesting and useful news for any kind of people in the society. Especially for the government agencies, it’s a very useful source as based on predictions, the harvesting and storing of water could be prepared well in advance. It also plays a key role in the production of power or generation power with the help of water flow in dams and reservoirs. In the article, an attempt has been made to build a system that takes input of previous years rainfall data from dataset which was collected from the Meteorology department of Andhra Pradesh and predict the average amount of rainfall of upcoming months in a specific year. We have initially separated the available data into training and testing data sets and made a model. We had applied various statistical and machine learning approaches like the linear model, Support Vector Machine algorithms etc. in predicting the results, along with a Neural Network model and made analysis over various approaches and compared their results with actual data. By these various approaches, the error of prediction was minimized and increased the accuracy of results predicted by the model.

Keywords:

Linear Regression, Support Vector Machine, Artificial neural network, rainfall forecasting, model selection