FORECASTING INFLATION IN NORTH SUMATRA PROVINCE USING ARIMA AND LINEAR REGRESSION MODELS

Abstract

This study compares the forecasting performance of the ARIMA and multiple linear regression models in predicting monthly inflation in North Sumatra Province, Indonesia. The dataset consists of 132 monthly inflation observations from January 2015 to December 2025 obtained from the Consumer Price Index (CPI) published by the Central Statistics Agency (BPS). A quantitative comparative approach was applied using training and testing datasets, where forecasting evaluation was conducted during the out-of-sample period from January to December 2025. Forecasting performance was assessed using RMSE, MAE, N-BIC, and R2. The results show that the ARIMA (1,0,1) model produced limited forecasting performance with an R2 of 6,9%, RMSE of 0,600, and MAE of 0,456. The multiple linear regression model incorpo rating time trends and monthly dummy variables achieved a relatively higher R2 of 16,3%, indicating better representation of seasonal inflation patterns. Overall, the findings provide exploratory empirical evidence for regional inflation forecasting.