基于稀疏QoS与协同过滤的个性化Web服务推荐方法研究
发布时间：2018-08-11 16:35
【摘要】：Web的快速发展带来了信息爆炸的现状,现在对个性化Web服务推荐信息系统的研究成为服务计算领域的一个热门研究方向。Web服务推荐系统的研究主要解决两个问题:稀疏QoS数据的预测及补全,用户个性化推荐。问题一是因为提供相同或相似功能的Web服务数量很多,对于个体用户而言,很有可能被推荐的服务是该用户从未接触过的,这样就存在QoS数据的稀疏问题,这就需要对缺失的数据进行补全。问题二是因为推荐给用户的服务是不是准确不能仅仅通过QoS值的预测来判断,还要考虑用户的个性化需求,在考虑用户的个性化需求的条件下对稀疏QoS值进行预测和补全,才能推荐合适的服务给目标用户。基于协同过滤的推荐系统能根据相似用户或相似服务的评分来预测当前用户的评分,在该研究方向已经有不少研究成果,但是QoS数据预测的准确性和个性化推荐的合理行方面仍然存在很多不足。在基于协同过滤的Web服务推荐算法研究方向,本文提出以下三种改进的算法模型:首先,本文提出了一种基于用户偏好的改进协同过滤Web服务推荐算法(UPCF),该算法的基本思路是,首先从QoS数据中提取用户偏好,并将其作为相似用户的选择标准,然后使用top-k算法确定目标用户及服务的相似邻居,最后使用调整的加权和方法来预测目标用户的QoS值。其次,本文在基于用户偏好的改进协同过滤Web服务推荐算法(UPCF)基础上,提出了基于联合用户偏好的改进协同过滤Web服务推荐算法(CUPCF)。该算法从QoS数据中提取用户偏好数据并使用于相似邻居的选择,在使用top-k算法确定目标用户及服务的相似邻居集合之后,使用QoS数据计算邻居的相似度,最后使用调整的加权和方法来预测目标用户的QoS值。最后,本文在CUPCF算法的基础上,提出了基于用户位置与偏好的改进协同过滤Web服务推荐算法(LSCUPCF),该算法将QoS数据按照用户的位置分布划分为几个子类,子类中的用户由于位置的相似具有更好的相似性,LSCUPCF在提高用户个性化考虑的基础上,降低了算法的计算复杂度,提高了算法的效率及准确率。我们的实验使用香港中文大学发布的WSDREAM数据集,该数据集收集了全球30个国家的339个用户和70多个国家的5825个Web服务,包含197万条真实Web服务QoS访问记录,WSDREAM数据集上的实验结果表明本文所提出的一系列推荐算法具有更好的预测准确率。
[Abstract]:The rapid development of Web has brought about the status quo of information explosion. Now the research on personalized Web services recommendation information system has become a hot research direction in the field of service computing. The research of web services recommendation system mainly solves two problems: sparse QoS data prediction and completion, user personalized recommendation. The first problem is that there are so many Web services that provide the same or similar functions that for an individual user, it is likely that the recommended service has never been in contact with that user, so there is a problem of sparse QoS data. This requires the completion of missing data. The second problem is that whether the service recommended to the user can not be judged by the prediction of the QoS value, but also by the consideration of the personalized demand of the user, and the sparse QoS value can be predicted and completed under the condition of considering the user's personalized demand. To recommend appropriate services to target users. The recommendation system based on collaborative filtering can predict the current users' scores according to the scores of similar users or similar services. However, there are still many shortcomings in the accuracy of QoS data prediction and the reasonable line of personalized recommendation. In the research direction of Web services recommendation algorithm based on collaborative filtering, this paper proposes the following three improved algorithm models: first, this paper proposes an improved collaborative filtering Web service recommendation algorithm based on user preference, (UPCF), the basic idea of the algorithm is: First, the user preference is extracted from the QoS data and used as the selection criterion for the similar users, then the top-k algorithm is used to determine the similar neighbors of the target user and the service. Finally, the adjusted weighted sum method is used to predict the QoS value of the target user. Secondly, based on the improved collaborative filtering Web service recommendation algorithm (UPCF) based on user preference, this paper proposes an improved collaborative filtering Web service recommendation algorithm (CUPCF). Based on joint user preference. The algorithm extracts user preference data from QoS data and uses it to select similar neighbors. After using top-k algorithm to determine the set of similar neighbors of target users and services, QoS data is used to calculate the similarity of neighbors. Finally, the adjusted weighted sum method is used to predict the target user's QoS value. Finally, based on the CUPCF algorithm, an improved collaborative filtering Web service recommendation algorithm based on user location and preference, (LSCUPCF), is proposed. The algorithm divides the QoS data into several subclasses according to the user's location distribution. The users in the subclass have better similarity because of the similarity of location. LSCUPCF reduces the computational complexity of the algorithm and improves the efficiency and accuracy of the algorithm on the basis of improving the personalized consideration of users. Our experiment uses the WSDREAM dataset released by the Chinese University of Hong Kong, which collects 339 users in 30 countries and 5825 Web services in more than 70 countries. The experimental results on the WSDREAM dataset containing 1.97 million real Web service QoS access records show that the proposed algorithms have better prediction accuracy.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3;TP393.09
[Abstract]:The rapid development of Web has brought about the status quo of information explosion. Now the research on personalized Web services recommendation information system has become a hot research direction in the field of service computing. The research of web services recommendation system mainly solves two problems: sparse QoS data prediction and completion, user personalized recommendation. The first problem is that there are so many Web services that provide the same or similar functions that for an individual user, it is likely that the recommended service has never been in contact with that user, so there is a problem of sparse QoS data. This requires the completion of missing data. The second problem is that whether the service recommended to the user can not be judged by the prediction of the QoS value, but also by the consideration of the personalized demand of the user, and the sparse QoS value can be predicted and completed under the condition of considering the user's personalized demand. To recommend appropriate services to target users. The recommendation system based on collaborative filtering can predict the current users' scores according to the scores of similar users or similar services. However, there are still many shortcomings in the accuracy of QoS data prediction and the reasonable line of personalized recommendation. In the research direction of Web services recommendation algorithm based on collaborative filtering, this paper proposes the following three improved algorithm models: first, this paper proposes an improved collaborative filtering Web service recommendation algorithm based on user preference, (UPCF), the basic idea of the algorithm is: First, the user preference is extracted from the QoS data and used as the selection criterion for the similar users, then the top-k algorithm is used to determine the similar neighbors of the target user and the service. Finally, the adjusted weighted sum method is used to predict the QoS value of the target user. Secondly, based on the improved collaborative filtering Web service recommendation algorithm (UPCF) based on user preference, this paper proposes an improved collaborative filtering Web service recommendation algorithm (CUPCF). Based on joint user preference. The algorithm extracts user preference data from QoS data and uses it to select similar neighbors. After using top-k algorithm to determine the set of similar neighbors of target users and services, QoS data is used to calculate the similarity of neighbors. Finally, the adjusted weighted sum method is used to predict the target user's QoS value. Finally, based on the CUPCF algorithm, an improved collaborative filtering Web service recommendation algorithm based on user location and preference, (LSCUPCF), is proposed. The algorithm divides the QoS data into several subclasses according to the user's location distribution. The users in the subclass have better similarity because of the similarity of location. LSCUPCF reduces the computational complexity of the algorithm and improves the efficiency and accuracy of the algorithm on the basis of improving the personalized consideration of users. Our experiment uses the WSDREAM dataset released by the Chinese University of Hong Kong, which collects 339 users in 30 countries and 5825 Web services in more than 70 countries. The experimental results on the WSDREAM dataset containing 1.97 million real Web service QoS access records show that the proposed algorithms have better prediction accuracy.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3;TP393.09
【参考文献】
相关期刊论文 前3条
1 许海玲;吴潇;李晓东;阎保平;;互联网推荐系统比较研究[J];软件学报;2009年02期
2 刘建国;周涛;汪秉宏;;个性化推荐系统的研究进展[J];自然科学进展;2009年01期
3 曾春,邢春晓,周立柱;个性化服务技术综述[J];软件学报;2002年10期
相关博士学位论文 前2条
1 孙慧峰;基于协同过滤的个性化Web推荐[D];北京邮电大学;2012年
2 高e，
本文编号：2177592
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2177592.html
