A novel retrieval strategy for case-based reasoning based on attitudinal Choquet integral

Retrieval is a very important stage in the case-based reasoning (CBR) process because it is the critical foundation for the success of the CBR system. Its goal is to retrieve valuable cases that can be employed as references to solve the target problem. The most commonly employed methods in the retrieval process today is similarity-based weighted average operators, which have been criticized for not considering the interactions among features. In this paper, we develop a novel retrieval strategy for CBR based on the attitudinal Choquet integral (ACI), which can capture (a) the features interaction, (b) relative features importance, and (c) the attitudinal character of decision maker. The core of the retrieval strategy is to define a global similarity which aggregates local similarity and feature similarity through ACI. In addition, to ensure the availability of data in the case base, we present a method of filling in missing data. The novel retrieval strategy and filling method are validated through two simulation experiments on several real data sets. The superiority of the developed approaches in terms of retrieval capability and filling efficiency can be demonstrated by approaching an average recognition rate of 82% and a filling accuracy of over 90%.