Use of Data Development Analysis for Certain Case Based Expert System Applications
暂无分享,去创建一个
Data Envelopment Analysis is a technique for comparing efficiencies of productive units based on their respective input and output data. The method is known as an efficient frontier technique and was developed originally by Charnes, Cooper, and Rhodes. This paper shows how the technique may be useful in an entirely different context namely that of case or example based expert systems. In this latter area, it is desired to make decisions, such as acceptance or rejection of credit risks, based on examples which have previously been decided by an expert. This paper shows how Data Envelopment Analysis (DEA) may be used to develop an acceptance boundary for use in case based expert systems. Acceptability of cases is identified with cases which lie on or above the efficient frontier in the DEA sense. The method requires convexity of the acceptable set to hold as its major condition. The method also assumes that tile accepted cases are accurately classified by the expert with respect to Type II errors.
[1] Robert W. Zmud,et al. A Synthesis of Research on Requirements Analysis and Knowledge Acquisition Techniques , 1992, MIS Q..
[2] Abraham Charnes,et al. Measuring the efficiency of decision making units , 1978 .
[3] Kurt Hornik,et al. Multilayer feedforward networks are universal approximators , 1989, Neural Networks.