This thesis aims at the design of expert system for stock portfolio selection. The focus of design is a synergetic combination of experts' knowledge and optimization models to supplement the shortcomings of purely knowledge-based expert systems and purely optimization-based models. Rule-based knowledge plays the role of screening unsatisfactory stocks and generating additional constraints in the optimization model.
The idea is realized by designing a system called KAIST (Korea Advanced Institute of Science and Technology) Portfolio Advisory System I which abbreviated KAPAS I. The major components of KAPAS I are the knowledge base, the model base and the data base.
Finally, an illustrative prototype is implemented in the microcomputer, and the issues for further research such as treatment of time dependent knowledge and uncertainty are identified.