This study proposes a method for combining three types of knowledge used by expert systems through fuzzy logic as an approach to improving the inferencing performance of expert systems. The types of knowledge used by expert systems can be broadly classified into user knowledge, expert knowledge, and machine knowledge acquired by computers. These types of knowledge possess different characteristics, and therefore, by combining them in a direction where synergy effects can be expected, the inferencing performance of expert systems can be improved even under uncertain conditions. The machine knowledge used in this study was obtained through a back-propagation artificial neural network model from empirical cases in the target problem domain, while the expert knowledge consists of knowledge representing trend changes in external environmental factors that influence the target problem domain. User knowledge represents the user's personal views regarding the information provided by expert knowledge and machine knowledge. To measure the effectiveness of the methodology proposed in this study, it was applied to the problem of predicting the trend of the Korean stock market one week ahead. The results statistically verified that the fuzzy logic-supported knowledge combination mechanism proposed in this study is highly useful for improving the inferencing capability of expert systems even in uncertain decision-making environments.