Rainfall Prediction using ARIMA and Linear Regression

Rainfall is the greatest of nature's gifts for our daily life, as well as the most important climate factor affecting human lives with farmers and agricultural complex systems. Rainfall forecasting is very difficult because excessive and sudden rainfall can have numerous problems, such as agriculture and property destruction, so for a better forecasting model is required for early warning that can save risk to agriculture and life property. Time series data have been used extensively in classical statistics. The proposed methodology predicts annual rainfall by time series ARIMA model and Linear Regression a machine learning algorithm. Time series data have been used extensively in classical statistics. The ARIMA has been trained to produce excellent outcomes. The ARIMA model demonstrated greater accuracy in all seasonal and yearly rains. To offer a solid prediction, this method, like time series ARIMA, requires a strict assumption of stationarity. We use real data from the Indian government website and Kaggle to compare model quality in ARIMA using different evaluation metrics. As a result, the ARIMA model accurately predicts rainfall and it is used for agriculture purpose in future.

1 Introduction

The rainfall analysis aids in the prevention and reduction of disasters such as droughts and floods. We can simply do it here by analyzing historical rainfall data and forecasting rainfall for future years. According to the needs, we may apply a different type of approaches such time series analysis and linear regression and it is used to calculate the error between the observed and forecasted values as well as the accuracy. Because by using different methods gives different values of accuracy which helps to select the bestmodel.