A novel text mining approach to financial time series forecasting

Financial time series forecasting has become a challenge because it is noisy, non-stationary and chaotic. Most of the existing forecasting models for this problem do not take market sentiment into consideration. To overcome this limitation, motivated by the fact that market sentiment contains some useful forecasting information, this paper uses textual information to aid the financial time series forecasting and presents a novel text mining approach via combining ARIMA and SVR (Support Vector Regression) to forecasting. The approach contains three steps: representing textual data as feature vectors, using ARIMA to analyze the linear part and developing a SVR model based only on textual feature vector to model the nonlinear part. To verify the effectiveness of the proposed approach, quarterly ROEs (Return of Equity) of six security companies are chosen as the forecasting targets. Comparing with some existing state-of-the-art models, the proposed approach gives superior results. It indicates that the proposed model that uses additional market sentiment provides a promising alternative to financial time series prediction.