Abstract:paper according to a rough set theory based on the method of the data mining, testing data with crossroads as the research object, by recording data to form the original decision table, and then to proceed data pre-processing the original decision table by improved Semi Naive Scaler algorithm. Finally, the decision table after the data preprocessing can be proceeded attribute reduction which is based on the discernibility matrix attribute frequency of heuristic reduction algorithm. The results obtained is the key attributes, which is the critical phase, then providing the basis for the path decision-making departments according to the program realization results.