提出了一种基于分类算法的潜在好友推荐系统. 该系统采用两步特征方法处理原始数据集，去除不相关特征项和冗余特征项，为分类器提供精简的特征集合；把学者潜在好友推荐问题转化为二分类问题，对比4个常用分类器在两步特征选择方法上的分类效果，并找出推荐效果最佳的分类器(决策树分类器),同时得出学术社交网络中区分度最大的6个用户特征信息. 使用来自学术社交网络学者网（SCHOLAT）的社交网络信息作为实验原始数据集进行测试，实验结果显示，相比传统方法，基于分类的推荐方法在准确率和F1值均有显著提升，体现了基于分类算法的潜在好友推荐系统的准确性和实用价值.
A potential friend recommendation system based on classification algorithm is proposed. The system uses the two-step feature method to process the original data set and removes irrelevant features for the classifier. The potential problems of scholars are translated into two classification problems. By comparing the classification effect of four commonly to find out the best classifier, the method of classifiers in two-step feature selection is used. At the same time, it concludes six influence factors and main diffusion characteristics. The social network information from the Academic Social Network (SCHOLAT) is tested as the original data sets. The experiment shows that the method based on significantly improves the accuracy and the F1 value. It reflects the accuracy and practical value of the recommendation system of potential friends based on classification algorithms.