面向空间相关性和加权评分效应的情境感知Web服务推荐算法研究

发布时间：2019-07-08 12:28

【摘要】：近年来,随着互联网迅速发展和移动设备的普及,互联网上出现了大量功能相似、种类繁多的Web服务。如何向用户推荐个性化的Web服务已成为服务计算领域的热点研究问题之一。传统的Web服务推荐算法已无法满足Web服务推荐的多元化需求。因此,考虑时间、空间等情境因素的情境感知Web服务推荐方法应运而生。然而,现有的情境感知Web服务推荐算法的研究仍存在如下问题:第一,现有算法主要关注于利用空间、时间等情境信息来寻找与当前用户的情境相似的用户或与当前Web服务的情境相似的服务。然而,尚未充分考虑用户情境和Web服务情境之间的关联性对用户Web服务偏好产生的影响因素。因此,Web服务的推荐结果难以响应用户或服务情境的动态变化。第二,在使用“用户-服务”Qo S值矩阵进行用户之间或服务之间相似度计算时,现有方案将大小不同的QoS值同等看待,而普遍忽略了极大或极小的QoS值对用户之间的相似度或服务之间的相似度的显著影响。因而,难以为用户推荐显著反映用户偏好的服务。针对上述问题,本文提出了一种面向空间相关性和加权评分效应的情境感知Web服务推荐方法(简称CASR-SCWRE算法)。该方法旨在为用户提供一种个性化的Web服务推荐机制:一方面,有利于挖掘用户情境和服务情境之间的相关性对用户Web服务偏好的影响,另一方面,保证推荐算法能够显著反映不同Qo S值对用户相似度或服务相似度的显著影响。本文的主要研究内容包括:首先,通过用户空间情境和服务空间情境之间的关联性,建模了空间相关性对用户Web服务偏好影响,得到与用户当前偏好相符的Web服务调用记录。其次,通过建模加权评分效应来计算用户之间相似度或服务之间相似度,并结合传统的时间衰减相似度求解模型,提出了基于加权评分效应的时间衰减模型。再次,基于上述两步过滤得到的数据,利用贝叶斯定理预测某一特定的Web服务对当前用户的QoS值。最后,本文在WS-Dream数据集上开展了一系列大规模的实验训练和测试。实验结果表明,CASR-SCWRE算法与诸多对比算法相比,具有更高的Web服务推荐精度。
文内图片：

图片说明：天气预报Web服务推荐场景图
[Abstract]:In recent years, with the rapid development of the Internet and the popularity of mobile devices, a large number of similar functions and a wide range of Web services have appeared on the Internet. How to recommend personalized Web services to users has become one of the hot research issues in the field of service computing. The traditional Web service recommendation algorithm can not meet the diversified requirements of Web service recommendation. Therefore, context-aware Web service recommendation method considering time, space and other situational factors emerges as the times require. However, the existing context-aware Web service recommendation algorithms still have the following problems: first, the existing algorithms mainly focus on using space, time and other situational information to find users who are similar to the current user's situation or similar to the current Web service situation. However, the relationship between user situation and Web service situation has not yet been fully considered, which affects the preference of user Web service. Therefore, the recommended results of Web services are difficult to respond to the dynamic changes of users or service situations. Secondly, when using the "user-service" QoS value matrix to calculate the similarity between users or between services, the existing schemes treat the QoS values with different sizes equally, and generally ignore the significant influence of the maximum or minimal QoS values on the similarity between users or between services. Therefore, it is difficult to recommend services that significantly reflect the preferences of users. In order to solve the above problems, a context-aware Web service recommendation method (CASR-SCWRE algorithm) for spatial correlation and weighted scoring effect is proposed in this paper. The purpose of this method is to provide users with a personalized Web service recommendation mechanism: on the one hand, it is beneficial to mine the influence of the correlation between user situation and service situation on users' Web service preference, on the other hand, it ensures that the recommendation algorithm can significantly reflect the significant influence of different Qo S values on user similarity or service similarity. The main research contents of this paper are as follows: firstly, through the correlation between user space situation and service space situation, the influence of spatial correlation on user Web service preference is modeled, and the Web service call record consistent with user's current preference is obtained. Secondly, the similarity between users or services is calculated by modeling weighted scoring effect, and combined with the traditional time attenuation similarity solution model, a time attenuation model based on weighted scoring effect is proposed. Thirdly, based on the data filtered in the above two steps, the QoS value of a particular Web service to the current user is predicted by using Bayesian theorem. Finally, a series of large-scale experimental training and testing are carried out on WS-Dream datasets. The experimental results show that the CASR-SCWRE algorithm has higher accuracy of Web service recommendation than many comparative algorithms.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09;TP391.3

【相似文献】