Author Login Editor-in-Chief Peer Review Editor Work Office Work

Computer Engineering

Previous Articles     Next Articles

Research on User Similarity Measure Method in Collaborative Filtering Algorithm

REN Kankan,QIAN Xuezhong   

  1. (School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)
  • Received:2014-08-11 Online:2015-08-15 Published:2015-08-15

协同过滤算法中的用户相似性度量方法研究

任看看,钱雪忠   

  1. (江南大学物联网工程学院,江苏 无锡 214122)
  • 作者简介:任看看(1986-),男,硕士研究生,主研方向:数据挖掘,数据库技术;钱雪忠,副教授。
  • 基金资助:
    国家自然科学基金资助项目(61103129,61202312);江苏省科技支撑计划基金资助项目(BE2009009)。

Abstract: User similarity measure method in traditional collaborative filtering algorithm is based on common items to calculate the similarity between users,and the low accuracy of the similarity calculation is caused by the data sparsity problem of user-item rating matrix.In view of this problem,a novel user similarity measure method is proposed.This method calculates the similarity of users by using Jaccard similarity coefficient which is improved by correction formula,considers the relationship between common items and all items of users during the calculating,and takes into account the impact of different of user ratings in the evaluation of common items on similarity of users,and more accurate user similarity matrix is obtained.Experimental results show that this method can improve the prediction accuracy compared with Cosine(COS) similarity method and the Adjusted Cosine(ACOS) similarity method,etc.

Key words: collaborative filtering, data sparsity, user similarity, Jaccard similarity coefficient, recommendation algorithm

摘要: 传统协同过滤算法中的用户相似性度量方法基于用户之间共同评分项计算用户的相似度,用户-项目评分矩阵的数据稀疏问题会导致该相似度的计算不够准确。为此,提出一种新的用户相似性度量方法。该方法采用结合修正公式改进的Jaccard相似性系数计算用户之间的相 似度,在计算过程中考虑用户之间共同评分项和所有评分项的关系,以及用户在共同评价项目上的评分差异对用户相似度的影响,从而获取更加精确的用户相似度矩阵。实验结果表明,与余弦相似性方法和修正的余弦相似性方法相比,该方法能提高预测准确度。

关键词: 协同过滤, 数据稀疏, 用户相似性, Jaccard相似性系数, 推荐算法

CLC Number: