A Hybrid DEA Based CHAID and Imperialist Competitive Algorithm for Stock Selection

This paper proposes a new framework for the formation of an optimal stock portfolio. The paper will argue that how an optimal stock portfolio is designed through the proposed approach compared with previous methods. In this paper, the investment portfolio is formed based on the data mining algorithm of CHAID on the basis of the risk status criteria. In the next step, the second investment portfolio is created based on the decision rules extracted by the DEA-BCC model. The final port-folio is created through a two-objective mathematical programming model based on the Imperialist Competitive algorithm. The proposed methodology is applied on a case study in the Tehran Stock Exchange. The results of the CHAID algorithm implementation based on the risk output field showed that all candidate stocks do not fall in one class and that is why it is necessary that each class of candidate stocks must be evaluated independently of other classes. The result of the Imperialist Competitive algorithm in small and large scale based on the Taguchi method showed that the studied stocks are calibrated with the used method. Unlike other models of stock portfolio selection, this paper first classifies the Stocks through the CHAID algorithm. The classified stocks in each class are evaluated independently of other classes through the DEA-BCC model. After narrowing the search space, the optimal portfolio is selected through the Imperialist Competitive algorithm.