Abstract

MODWT-ARIMA is a time series modeling that combines the MODWT process and the ARIMA process. The MODWT process is used as pre-processing data while the ARIMA process as a time series modeling for data from MODWT decomposition. This study aims to show that time series modeling with a combined MODWT-ARIMA process provides more accurate forecast result compared to the ARIMA model. The modeled data is time series of daily stock price BBRI.JK started from January 2, 2015 to December 31, 2018. Accuracy measurement of the forecasting result is based on the RMSE value. The result is the MODWT-ARIMA model has a RMSE value  which is smaller than the ARIMA model with RMSE , while the RMSE forecast results  for 43 future periods is  which is also smaller than the ARIMA forecast RMSE, . The diagnostic checking results if the ARIMA model for MODWT decompotition data, namely D1, D2, D3, and S3, indicate that the residual model is not white noise, while the ARIMA model for the time period of daily stock prices has white noise residuals. Theoritically, a model that has no white noise’s residual is considered to be less able to describe the properties of the observed data and further residual modeing should be done. However, this research is sufficient for the ARIMA model and it can be shown that the MODWT-ARIMA model is more effective for modelling time series that are not stationery compared to the ARIMA model.