推薦系統是數據挖掘中強有力的技術，異構信息網絡是推薦系統起步晚卻發展迅猛的主流推薦方法。提出了基于異構信息網絡的學生成績預測與預警模型，該方法通過元路徑計算得到學生間相似度矩陣，利用相似度矩陣構造成績變化趨勢矩陣和幅度矩陣，投票得到學生成績預警與預測結果；最后，在公開數據集上驗證所提模型的有效性，結果表明，該模型能夠對學生成績進行預警，并能在一定閾值下預測學生成績具體分值。

Research on the prediction and early warning model of student achievement based on heterogeneous information network

Abstract： Abstract： Recommendation system is a powerful technology in data mining. Heterogeneous information network is the mainstream recommendation method which started late but developed rapidly. In this paper, we proposed a student achievement prediction and early warning model based on heterogeneous information network. This method obtains the similarity matrix among students by Meta-Graph calculation, constructs the achievement change trend matrix and amplitude matrix by using the similarity matrix, and obtains the student achievement early warning and prediction by voting. Finally, the effectiveness of the proposed model is verified on the open data set, the experimental results show that the model can give early warning to the student achievement and can predict the specific score of students′ performance under a certain threshold.

Key words : heterogeneous information network；Meta-Graph；academic warning；academic prediction；data mining