It has not satisfied general investors’ needs by simply saving their money into account or investing their money in stock market, when they have begun to search many distinct ways to manage their money. In the long-term, depending on its major advantages which risk is smaller than stock market and rate of return is higher than the money that save into account, the mutual funds have become an indispensable tool of managing money these days. However, evaluating their performance is an important issue while facing numerous mutual funds. As to the research on the performance of the mutual funds so far, it mostly has been measured with financial model、general statistical techniques and value at risk. Many of previous researches haven’t used data mining as research ways, so this study puts emphasis on whether it has co-movement among the stock funds of different risk type through data mining. Furthermore, it belongs to positive or negative comovement if it has co-movement. This study would provide investors with a new thought and research technique which is totally different from previous ways. Finally, the technique of data mining would be used to select stock funds in this study, taking advantage of association rule to find some hidden rules and making a realistic portfolio for general investors; moreover, the realistic portfolio would be proved that its performance is superior to the market through the analysis of return and risk. In addition, this study would also stand from a corporation’s angle, devising an umbrella funds and advising different percentage of investment for diverse attribute of investors.