推荐算法背景下兴趣社区挖掘算法

Interest Community Discovery Algorithm in Recommendation Network

针对传统的网络拓扑构建方法难以很好描述推荐平台中网络用户的情况,提出了一种基于用户“围观”行为的网络构建方法,对构建的网络进行分析。由于传统的标签传播算法易发生标签入侵和算法结果不稳定,改进标签传播算法的标签更新规则,提出一种适用于挖掘推荐网络社区的标签传播算法。利用短视频推荐平台数据集,对所提算法进行了实验论证。结果表明,所提算法挖掘的社区内部兴趣相似度显著高于整个网络中的兴趣相似度,有较高的兴趣内聚性。

The traditional network topology construction method in the recommendation platform was difficult to describe the network users.To solve this problem,a network construction method based on the“click”behavior of users was proposed,and the constructed network was analyzed.Aiming at the shortcomings that the traditional label propagation algorithm was prone to label intrusion and unstable algorithm results in the recommendation network,an algorithm that changed the tag update rules of the tag propagation algorithm was proposed,which was suitable for mining recommended network communities.An experiment was conducted using the domestic short video recommendation platform data set.The results showed that the interest similarity within the community mined by the proposed algorithm was significantly higher than that in the whole network,and had high interest cohesion.