For the selection of stock portfolio, the optimization models or expert systems are utilized separately. To take advantage of both approaches, the integration of optimization model and expert system is attempted. The proposed architecture - ISPMS (Intelligent Stock Portfolio Management System) - accomplishes the integration by interpreting the knowledge as a part of the formulation of optimization model. The other generic issues in the ISPMS are the integration of personal preference with the expert``s knowledge, the knowledge acquisition by machine learning, and the conflict resolution among knowledge from diverse sources. This thesis thus describes the representation and the inference of an expert``s knowledge: the representation of personal preference and its integration with the expert``s knowledge: the knowledge acquisition by machine learning: the conflict resolution among knowledge: the interpretation of the knowledge and preference to associate with optimization model: and solution algorithm of modified optimization model accordingly. To support the concept of ISPMS, the proposed prototype named K-FOLIO has been implemented using Common LISP and Turbo Pascal in the super microcomputer environment.