When we use Case-Based Reasoning (CBR) for practical applications, it is often the case to implement two kinds of problem solvers: casebased one and knowledge-based/conventional one. The complex integration of such plural reasoners often causes serious VV~T problems: performance of problem solving, quality of solutions, and so on. In this paper, we ~ldress the performance validation problems of a CBR system. This paper presents a blackbox validation approach to quantitatively analyze the performance and proposes three kinds of performance measures. The proposed measures are then applied to validating a CBR system: IRS-CHR (Intelligent Information Retriever with a Case-Based Reasoner). From the experimental results, we conclude that (1) IRS-CBR has succeeded in both speed-up and memory-based learning, and (2) the proposed measures are useful for validating a CBR system.
[1]
Kristian J. Hammond,et al.
Case-Based Planning: Viewing Planning as a Memory Task
,
1989
.
[2]
Takao Terano,et al.
The JIPDEC checklist‐based guideline for expert system evaluation
,
1994,
Int. J. Intell. Syst..
[3]
Christopher K. Riesbeck,et al.
Inside Case-Based Reasoning
,
1989
.
[4]
Hee-Youn Lee,et al.
Prototype-Based Reasoning: An Integrated Approach to Solving Large Novel Problems
,
1991,
AAAI.
[5]
Paul S. Rosenbloom,et al.
Improving Rule-Based Systems Through Case-Based Reasoning
,
1991,
AAAI.