基于聚类系数的推荐算法*

摘 要：针对于标准二分图网络推荐算法（NBI）的物质扩散机制过于简单的问题，基于聚类系数的改进NBI算法（简称NBICC）被提出。在文章中，推荐系统可以被抽象为一个有向加权二分图网络。在物质扩散的过程中，考虑到聚类系数因素的影响，重新定义了商品之间的相似度的计算公式，进而获得了更加精确的推荐结果。Ranking score,precison,recall等评价指标被应用在本文提出的新算法中，实验结果表明，这三样重要指标上，NBICC算法都强于标准NBI算法。

A Modified Recommendation Algorithm Based on Clustering Coefficient

Abstract: Accordance with the problem that the mass diffusion mechanism which standard NBI algorithm used is too simple, a modified NBI algorithm based on clustering coefficient (NBICC) is proposed. In this article, recommendation system is regarded as a direct graph with weight. In order to obtaining more accurate result, the calculation formula of similarity was redefined by considering clustering coefficient in the process of mass diffusion. Numerical results indicating that the algorithmic accuracy, measured by the average ranking score, precision and recall is improved greatly.