Facing with the optimization problem of fuzzy rules in the fuzzy time series forecasting model, an algorithm is proposed to optimize fuzzy rules by combing auto-correlation theory with fuzzy time series. Firstly, data stationarity is discussed and then fuzzy sets are obtained by using the traditional data fuzzification method, thereby fuzzy rules are established. Fuzzy rules are optimized by using the auto-correlation theory. Finally, through the forecasting of Alabama university enrollments, results show the effectiveness of the proposed method.