基于项目类别和兴趣度的协同过滤推荐算法  

Collaborative filtering recommendation algorithm based on item category and interest

作　　者：韦素云[1] 业宁[1] 吉根林[2] 张丹丹[1] 殷晓飞[1] 

机构地区：[1]南京林业大学信息科学技术学院,南京210037 [2]南京师范大学计算机科学与技术学院,南京210046

出　　处：《南京大学学报（自然科学版）》2013年第2期142-149,共8页Journal of Nanjing University（Natural Science）

基　　金：国家"973"项目(2012CB114505);国家杰出青年基金(31125008);江苏省自然科学基金(BK2009393);江苏省青蓝工程(CXLX11_0525);南京林业大学科技创新项目(163070079);江苏高校大学生创新计划项目(164070742)

摘　　要：用户评分数据极端稀疏情况下,传统相似性度量方法存在弊端,导致推荐系统的推荐质量急剧下降.针对上述问题,提出一种基于项目类别和兴趣度的协同过滤推荐算法.在该算法中,首先通过计算项目之间的类别距离,构造项目类别相似性矩阵;然后采用兴趣度分析不同项目之间的相关程度;最后结合项目类别信息和项目间的兴趣度,使用改进的条件概率方法作为衡量项目间相似性的标准.实验结果表明,该算法可以有效缓解用户评分数据稀疏带来的不良影响,提高预测准确率和推荐质量.Collaborative filtering-based recommender systems,which automatically predict preferred products of a user using known preferences of other users,have become extremely popular in recent years due to the increase in web-based activities such as e-commerce and online content distribution.However,traditional collaborative filtering techniques provide poor accuracy,a large number of ratings from similar users or similar items are not available,due to the sparsity inherent to rating data.Consequently,prediction quality can be poor.To address the matter,a new collaborative filtering recommendation algorithm based on item category and interest measure is proposed.In this algorithm,first,the item categories similarity matrix is constructed by calculating the item-item category distance,and then analyzes the correlation degree of different items by using Piatetsky-Shapiro interestingness measure,at last,a novel collaborative filtering algorithm is proposed after combining the information of item categories with item-item interestingness and utilizing ameliorated conditional probability method as the standard item-item similarity measure.Empirical evaluation of the algorithm on large movie rating datasets demonstrates that it is not only an effective solution to data sparisity and the drawbacks of traditional similarity method,but also improves the accuracy of user interest and nearest neighbor search.At the same time,this algorithm achieves better prediction accuracy compared to other well-performing collaborative filtering algorithms.

关 键 词：推荐系统 协同过滤 项目相似性 项目类别相似性 项目兴趣度 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]