Weekly Temperature Prediction by SARIMA Model in Central India

Abstract

A study was carried out to develop a Seasonal Autoregressive Integrated Moving Average (SARIMA) model for weekly temperature prediction at Tikamgarh district of central India. Weekly maximum and minimum temperature from 1980 to 2018 were utilized for training and from 2019 to 2021 for the testing purpose. High accuracy SARIMA models were selected for weekly maximum and minimum temperature prediction and the analysis carried out for that are discussed in this paper. The accuracy of the models was tested through the autocorrelation function and the partial autocorrelation functions. It was found that the SARIMA (1,0,0)(1,1,0)₅₂ and SARIMA(1,0,1) (2, 1, 0)₅₂ model performed better results. The MAE values are 1.81, 1.46, RMSE of 2.38, 1.93 and ACF1 value 0.003, 0.013 for SARIMA models. The selected models were showed ability for prediction of weekly maximum and minimum temperatures with accuracy. The analysis showed that the through SARIMA approach weekly temperature prediction with accuracy may be achieved in central India.