The objective of this paper is to propose a knowledge-based fuzzy post adjustment so that unstructured problems can be solved more realistically by expert systems. Major part of this mechanism forcuses on fuzzily assessing the influence of various external factors and accordingly improving the solutions of unstructured problem being concerned. For this purpose, three kinds of knowledge are used : user knowledge, expert knowledge, and machine knowledge. User knowledge is required for evaluating the external factors as well as operating the expert systems. Machine knowledge is automatically derived from historical instances of a target problem domain by using machine learning techniques, and used as a major knowledge source for inference. Expert knowledge is incorporate dinto fuzzy membership functions for external factors which seem to significantly affect the target problems. We applied this mechanism to a prototyoe expert system whose major objective is to provide expert guidance for stock market timing such as sell, buty, or wait. Experiments showed that our proposed mechanism can improve the solution quality of expert systems operating in turbulent decision-making environments.