Applying a hybrid data envelopment analysis approach to construct an intelligent stock trading system

The predictability of stock market returns has been investigated in the financial literature for over four decades. Various prediction or stock classification tools, such as conventional statistical methods and non-parametric methods, have been successfully developed in the literature to enhance the existing financial analysis. Among these tools, the DEA (data envelopment analysis)-based stock classification method is the one received most discussion lately. However, literature has shown that the efficiency estimates of a stock obtained from the slack variable analysis can be biased if the inputs/outputs of this stock are strongly correlated. In this paper, a two-stage approach of integrating independent component analysis (ICA) and data envelopment analysis (DEA) is proposed to overcome this issue. We suggest using ICA first to extract the variables for generating independent components, then selecting the ICs to represent the independent sources of variables, and finally, inputting the selected ICs as new variables in the DEA model. A financial data on the selected integrated circuit companies collected from the Taiwan Economic Journal and the Taiwan Stock Exchange Corporation (TSEC) during 2004 to 2006 is used to demonstrate the validity of the proposed two-stage approach. The results show that the proposed method has not only the highest classification accuracy but also the lowest potential loss. Hence the proposed approach can reduce the possible high risks in stock selection.