Comparison and Forecasting for Indian Rainfall Using Proposed and Time Series Models

Abstract

The experiment was conducted using rainfall data for the period from 1968 to 2022 in India. The aim of this study was to evaluate and compare the effectiveness of time series and proposed models in accurately predicting seasonal and annual rainfall trends in India. Since Indian agriculture was highly dependent on rainfall, this study attempted to propose a method for forecasting rainfall in India. Firstly, the initial step-1 involves specifying both the linear and multiplicative regression models with available independent variables with and without one time period log dependent variable as one of the independent variables and select one of the above four specifications having maximumfor forecasting purpose. In step-2 that all of the model’s independent variables, such as linear, exponential, and power functions, are correlated with a time variable. Estimate these relationships for every independent variable, choose the best one based onvalues, and predict the future values for all independent variables. As a step-3 substitute the future values of the independent variables obtained from step-2 in the selected model obtained from step-1. The obtained values are called forecasted values through proposed method. As a step-4, the dependent variable was related to time as linear, exponential and power functions; select one of these estimated relations with maximumfor forecasting. Finally, to compare the forecasted values through proposed and time series models usingcriteria. The results show that the proposed forecasting model was better than a time series forecasting model. Since the proposed models are the best models.