基于不平衡样本推荐算法研究

Research on Algorithm of Recommendation Based on Unbalanced Samples

用户或项目的分类是推荐算法的关键内容,而推荐过程中的不平衡样本问题,会影响少数类用户或少数类项目的推荐效果。笔者提出了SVM集合算法,利用SVM对多数类样本和少数类样本两种不同样本凸包分别进行压缩,而不会对SVM的分类超平面造成影响的几何特性,来实现不平衡样本的分类效果。

User or project classification is a key element of the recommended algorithm,and the unbalanced sample problem in the recommended process can affect the recommendation of a few categories of users or a few categories of projects.In this paper,SVM ensemble algorithm is proposed,and SVM is used to compress the two different sample convex hulls of most samples and few samples,and the geometric characteristics of SVM are not affected by the classification of SVM.The classification effect of unbalanced samples is realized.