Multivariate analytical knowledge acquisition method for knowledge based planning system

The authors propose a new knowledge acquisition method for solving planning problems. Since original teaching data include ambiguous and faulty data, the data distribution is first evaluated by discriminant analysis. Then, by introducing confidence probability and suspended region concepts, effective data are selected and used as teaching data for the version space method. The proposed method can acquire strategic knowledge in concise descriptions and has been successfully installed in several scheduling and operation assignment systems. These applications confirm a reduction in the knowledge acquisition effort