INTEGRATION OF HUMAN KNOWLEDGE AND MACHINE KNOWLEDGE BY USING FUZZY POST ADJUSTMENT - ITS PERFORMANCE IN STOCK-MARKET TIMING PREDICTION

This paper proposes a fuzzy post adjustment (FPA) mechanism so that hu man knowledge and machine knowledge can be integrated more synergistic ally to improve the performance of expert systems. Machine knowledge m eans knowledge algorithmically derived from past instances. Human know ledge implies (1) expert knowledge judging the trends of external fact ors and (2) user knowledge representing users' personal views about in formation given by both expert knowledge and machine knowledge. We con sider an expert system that uses the FPA mechanism to incorporate the effect of external factors effectively into ifs inference process. The goal of this expert system is stock market timing prediction, which i s divided into four kinds: bull, edged-up, edged-down and bear. Empiri cal tests showed that the proposed FPA mechanism can improve the perfo rmance of an expert system significantly, even in a turbulent decision -making environment.