Weighted SVM-ARIMA hybrid model for financial time series forecasting

With the increasing importance of forecasting with the high degree of accuracy, many forecasting approaches have been broadly developed to forecast in an ccurate way. Series hybrid methodology is one of the most commonly-used hybrid approaches that has encountered a great amount of attractiveness in the literature of time series forecasting and has been applied successfully in a wide variety of domains. However, conventional series hybrid models proposed in the literature are established based on the decomposing time series into linear and nonlinear parts and generating linearnonlinear modeling order for modeling decomposed components. Another assumption considered in the traditional series model is assigned equal weights to each model used for modeling linear and nonlinear components. Thus, in this paper in contrary to traditional series hybrid models, in order to improve the performance of series hybrid models, these two basic assumptions have been violated. The main aim of this study is to propose a novel weighted SVM-ARIMA model filling the gap of series hybrid models by changing the order of sequence modelling and assigning appropriate weights for SVM and ARIMA employing Ordinary Least Square (OLS) weighting algorithm. The effectiveness of the proposed hybrid model is proved through the benchmark data sets e.g. Dow Jones Industrial Average Index (DJIAI) and Nikkei 225 (N225) stock price. The experimental results verified that the proposed model outperforms the ARIMA-SVM, ARIMA, and SVM models in stock price forecasting.

کلیدواژه ها:

Series Hybrid Model ، Weighted SVM-ARIMA model ، Auto-Regressive Integrated Moving Average (ARIMA) ، Support Vector Machine (SVM) ، Financial Time Series Forecasting