Feature selection in Pattern recognition and Data mining, and other fields have a very wide application .Comparing a variety of feature selection methods in the practical application of the Uyghur text categorization, this paper uses CHI statistical method for feature selection, due to the traditional CHI statistic is calculated only shows the entry of contribution degree, but cannot explain entries and the categories of correlation. Therefore, this paper improved the CHI statistic, and combined with Pattern aggregation method, so as to realize the Uyghur text feature selection. Experimental results show that the proposed approach in this paper can not only save the time of feature selection and classification, but also greatly improve the quality of the classification.