A case retrieval method combined with similarity measurement and DEA model for alternative generation

Abstract

In alternative generation, reusing past experience is a potential methodology and case retrieval is a primary step. In order to improve the performance of case retrieval process, many applications have used different similarity measurements and the selection method for the most suitable historical case to solve problems. Many investigations have shown that human beings are usually bounded rational and their psychological behavior has certain influence on decision making. However, such behavior is neglected in similarity measurements and the selection method can only deal with the evaluation given by one decision maker (DM). This paper proposes a new case retrieval method that combines similarity measurement and data envelopment analysis (DEA) model. A similarity measurement based on cumulative prospect theory is proposed to consider the DM’s psychological behavior. A hybridization of four similarity measurements is used to generate a set of similar historical cases. The DM evaluates the similar historical case set by a pairwise comparison matrix. A DEA model is constructed to get the priority vector. The most suitable historical case can then be picked out through the case similarity and the case priority. A case study is finally introduced to illustrate the use of the proposed method.