摘要
协同过滤是推荐系统中广泛使用的一种推荐技术,但是目前多数基于协同过滤的Web服务推荐算法默认用户反馈数据是可信的,没有考虑到用户反馈数据中会出现的一些范围异常数据和恶意评价数据。如果没有对含有这些干扰数据的用户进行处理,最终的预测结果会受到严重的影响。在分析了用户反馈的服务质量(Qo S)数据的基础上,提出了一种基于用户可信度的Web服务推荐方法。计算用户的可信度并根据可信度对用户进行聚类,选取可信度高的类别中的数据进行协同过滤完成推荐,排除了范围异常数据以及恶意评价数据的干扰,避免了对预测结果的影响,最后使用平均绝对误差和均方根误差评判预测结果。实验结果表明,该方法能有效提高Qo S预测的准确率和Web服务的推荐质量。
Collaborative filtering is widely used in recommended systems.However,most of the existing Web service recommendation algorithms based on collaborative filtering assume that the user feedback is credible and do not take into account some range-abnormal data and malicious evaluation data in the user feedback,which will bring serious impact on the final prediction results.By analyzing the quality of service（ Qo S） data,we propose a Web service recommendation method based on user credibility. Specifically,the credibility of each user is calculated firstly,and then the users are grouped into several clusters according to their credibility.The data in the cluster with high reliability are selected to perform the collaborative filtering.In this way,the influence of the range-abnormal data and the malicious evaluation data to the prediction result can be decreased.Finally,the average absolute error and the root mean square error are used to evaluate the prediction result.The experiment shows that the proposed method can effectively improve the accuracy of Qo S prediction and the recommended quality of Web services.
作者
曹继承
朱小柯
荆晓远
吴飞
CAO Ji-cheng;ZHU Xiao-ke;JING Xiao-yuan;WU Fei(Schoolof Computer,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;State Key Laboratory of Software Engineering,School of Computer,Wuhan University,Wuhan 430072,China;School of Automation,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)
出处
《计算机技术与发展》
2018年第7期117-120,124,共5页
Computer Technology and Development
基金
国家自然科学基金(61272273)
江苏省自然科学基金(BK20170900)
南京邮电大学引进人才科研启动基金(NY217009)
关键词
协同过滤
Web服务推荐
用户可信度
聚类
collaborative filtering
Web service recommendation
credibility of users
clustering
