The support vector regression (SVR) model is a novel forecasting approach and has been successfully used to solve time series problems. However, the applications of SVR models in a seasonal time series forecasting has not been widely investigated. This study aims at developing a seasonal support vector regression (SSVR) model to forecast seasonal time series data. Seasonal factors and trends are utilized in the SSVR model to perform forecasts. Furthermore, hybrid genetic algorithms and tabu search (GA/TS) algorithms are applied in order to select three parameters of SSVR models. In this study, two other forecasting models, autoregressive integrated moving average (SARIMA) and SVR are employed for forecasting the same data sets. Empirical results indicate that the SSVR outperforms both SVR and SARIMA models in terms of forecasting accuracy. Thus, the SSVR model is an effective method for seasonal time series forecasting.
The seasonal time series is a sequence of seasonal data points recorded sequentially in time. Over the past several decades, many works have been devoted to develop and improve seasonal time series forecasting models. The SARIMA model (Box & Jenkins, 1976) is one of the most popular approaches in seasonal time series forecasting. The SARIMA model has been successfully utilized in many fields of forecasting, such as a soil dryness index (Li, Campbell, Haswell, Sneeuwjagt, & Venables, 2003), predicting tourism demand (Goh and Law, 2002 and Huang and Min, 2002) and municipal solid waste management (Navarro-Esbrı´, Diamadopoulos, & Ginestar, 2002). However, the seasonal time series is a complex and nonlinear problem. The artificial neural network (ANN) model is an alternative in forecasting seasonal data pattern (Zhang & Qi, 2005). Some literature (Nam and Schaefer, 1995, Pai and Hong, 2005, Tang and Fishwick, 1993 and Williams, 1997) indicted that ANN can obtain desirable results in seasonal and trend forecasting. With the introduction of Vapnik’s ε-insensitive loss function, SVR ( Vapnik, Golowich, & Smola, 1996) has been extended so as to solve forecasting problems and has provided many exciting results. In recent years, SVR schemes have been extended to cope with forecasting problems, and have provided many promising results in customer demand ( Levis & Papageorgiou, 2005), finance ( Huang et al., 2005, Kim, 2003 and Tay and Cao, 2002) intermittent demand ( Hua & Zhang, 2006), tourism demand ( Pai & Hong, 2005), air quality ( Lu & Wang, 2005), wind speed ( Mohandes, Halawani, Rehman, & Hussain, 2004), plant control systems ( Xi, Poo, & Chou, 2007), rainfall ( Hong & Pai, 2007), prices for the electricity market ( Gaoa, Bompard, Napoli, & Cheng, 2007) and flood control ( Yu, Chen, & Chang, 2006). However, applications of the SVR models in seasonal time series data have not been widely studied. Therefore, this study attempts to develop a SSVR model for exploiting the unique strength of the decomposition techniques and for the SVR model in the seasonal time series forecasting problems. The rest of this paper is organized as follows. The SSVR model is introduced in Section 2. In Section 3, two numerical examples are utilized to demonstrate performances of different forecasting models. Finally, conclusions are made in Section 4.
SVR models have been successfully used in time series forecasting problems, but they have not been widely explored in seasonal time series prediction. This study proposed a hybrid SSVR model for forecasting seasonal time series data, with experimental results being valid and satisfied. The superior performance of the SSVR model can be ascribed to two causes. First, the decomposition method enhances the ability of SSVR models in capturing seasonal nonlinear data patterns. Second, the GA/TS algorithm properly provides three parameters for the SSVR model. For future work, forecasting other types of time series data by a SVR-related model is a challenging issue for study. Another future research direction could consider using data preprocessing techniques for improvement in the forecasting accuracy of the SSVR model in a seasonal time series data.
