Knowledge refinement approach through incorporating case-based knowledge in maintenance engineer scheduling AI system

Monthly scheduling for elevator maintenance engineers is not only as complex as a large scale travelling salesman problem of several hundreds cities, but also involves more than 100 complex, exceptional and rather emotional conditions such as customer satisfaction and working conditions. An AI system for this scheduling was developed and has been practically used for more than seven years. However, above-mentioned conditions are different monthly, yearly among each of hundreds of maintenance bases in the whole country. Therefore, knowledge refinement cost is quite expensive compared with that needed for ordinary software systems or simple expert systems. In this paper, the problems of the knowledge refinement in this scheduling system are analyzed, and to overcome these problems, two (naive/elaborate case-based) approaches incorporating case-based knowledge are proposed for efficient knowledge refinement of a practical complex AI scheduling system. Experts mostly accepted schedules generated by knowledge refined through either naive or elaborate case-based approaches, though they had been seriously complaining about those generated by knowledge refined for about 4 years without using these approaches