基于三部图的随机游走知识推送方法研究

Study on Knowledge Push Method Based on Tripartite Graphs Random Walk with Restart

针对传统知识推送方法中数据稀疏性的问题,本文提出了一种基于三部图的随机游走知识推送方法。该方法首先建立并分析了"用户-项目-标签"三部图,得到了用户对项目和标签的初始评分值;然后分别在项目空间和标签空间中利用随机游走算法,生成若干个待推送项目,并重新计算预测评分;最后对用户进行知识推送。实验结果表明,该推送方法有效地提高了知识推送的精确度,满足了用户的知识需求。

This paper proposes a random walk knowledge push method based on tripartite graphs, to solve the problem of data sparsity in traditional knowledge push method. This method builds a ＂user-project-tag＂ tripartite graphs, and generates users＂ initial values of projects and tags, then uses random walk algorithm in the project space and tag space to generate several projects pending to be pushed, and refre- shes the values, finally pushes them to users. The experimental results shows, this method can improve the accuracy of knowledge push ef- fectively to satisfy users＂ demands.