Utilizing the Hybrid Models of Double Exponential Smoothing and Double Moving Average with Fuzzy Time Series Markov Chain for Stock Price Forecasting

Abstract

Double exponential smoothing (DES) and double moving average (DMA) are forecasting algorithms that are well-suited for time series data with trend patterns as they can be used to quickly identify the trend direction by smoothing the data. However, both models tend to give late signals and can only capture linear relationships in the data. Therefore, they should be combined with other models that are effective in nonlinear modeling, such as fuzzy time series Markov chain (FTSMC), for forecasting financial time series data, such as stock prices, that have not only linear relationships but also nonlinear relationships. This paper proposed the use of hybrid approaches of DES-FTSMC and DMA-FTSMC models to analyze the trend time series data. The models are combined in order to effectively capture different forms of patterns in the data. The hybrid models utilize DES and DMA to identify the direction of the trend and FTSMC to model the residual series after the removal of the trend effect. The proposed hybrid approaches are applied to the daily closing stock price of PT Indosat Ooredoo Hutchison Tbk. The results show that the DES-FTSMC hybrid model generates a MAPE value of 1.09% on the training data and 0.89% on the testing data. While the DMA-FTSMC combination yields MAPE values of 1.24% and 1.19% on the training and testing data respectively. This finding suggests that the proposed hybrid DES-FTSMC model is the better forecasting model for achieving higher accuracy.