Which is More Reliable, Expert Experience or Information Itself? Weight Scheme of Complex Cases for Health Management Decision Making

Abstract

How to obtain valuable knowledge more effectively from historical cases and satisfy the requirements of supporting diagnosis or management decision making is one of the important and challenging issues in the research field of modern historical information management and intelligent decision-making science. In this study, we develop a novel case-based reasoning (CBR) method which is based on information entropy and improved gray systems theory for knowledge acquisition of historical diagnosis decision-making cases. Specially, information entropy for weight determination is introduced into the CBR, as well as a gray system theory combined to support the diagnosis decision making of breast cancer. Based on two different real-world data sets, we conduct experimental studies to compare the performance of the Delphi method and information entropy. We also investigate which combination is best among different weight determination methods and retrieval algorithms. The results suggest that: generally, information entropy is a better approach to weight derivation and better matching effect can be obtained if it is integrated into the retrieval algorithm based on gray system theory rather than Euclidean distance algorithm. Our study can provide a novel approach to obtain weight values of cases, as well as an effective tool to mine valuable decision knowledge from historical cases in public hospitals.