Graduate Adjusting is an important step for Graduate Admission. The traditional adjusting methods which are all manual, make it very hard for students to choose a proper school from a huge number of data. This paper proposes a data-mining method based on semi-supervised study. Using the association rules, which are extracted from the labeled training samples, as supervised information, and combining with the K-mean algorithm in non-supervised study method, this paper elaborates on the semi-supervised study algorithm by classifying a large number of unlabeled data. This method overcomes the defects of inaccuracy in traditional methods which are influenced by a large number of factors. The method is simple to implement, has high accuracy, and can be widely used.