Abstract:
The objective of this research is to study and compare two models of forecasting time series, namely multiplicative seasonal ARIMA model and decomposition model with ARIMA. In this study, we use the monthly time series data of residential electricity consumption in Thailand. The forecasting time series were separated into two duration times. The first one is the forecasting time series data in previous time from January 2002 to December 2015 for 168 months. This data was used for selection the suitable models in each forecasting models. As the second one is the lead forecasting data from January to December 2016 for 12 months, which used for selection the most suitable model. The Mean Absolute Percentage Error (MAPE) and R-Squared are used as the comparative criteria. It was found that the decomposition model with ARIMA is the most suitable model for forecasting the monthly time series data of residential electricity consumption in Thailand. The decomposition model with ARIMA for 12 months lead forecast had the lowest MAPE of 3.0590 percent with the highest R-Squared of 98.8303 percent.