时空上下文感知的云服务质量预测和推荐的研究

项目名称： 时空上下文感知的云服务质量预测和推荐的研究

项目编号： No.61502212

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 谢春丽

作者单位： 江苏师范大学

项目金额： 20万元

中文摘要： 为解决传统二维数据下服务QoS预测精度不足的问题，本项目给出了基于三维张量分解的解决方案。首先从服务的质量感知数据、用户评价数据以及用户反馈数据中抽取出QoS感知数据的时间和空间上下文信息，构成用户-服务-时间（空间）形式的三维QoS数据；其次，由于一般用户仅对少数云服务有过使用记录，导致得到的时空上下文质量感知数据中有大量的未知值，本项目将云服务的质量预测问题转化为张量分解问题并求解。再次，为解决QoS的稀疏性问题，提出基于时间和空间相似的张量分解模型。最后，构建本项目的实验平台，将时间和空间上下文预测模型进行融合，构建多元QoS预测模型。本项目为用户选择满足QoS要求的云服务提供了基础，并为服务提供者提供了将服务推荐给潜在客户的功能。

中文关键词： 服务质量预测；时间感知；空间感知；云服务；服务推荐

英文摘要： To improve the traditional QoS prediction accuracy which was based on the two dimension data, the project gives a solution based on three dimension tensor factorization model. Firstly, the project extracts the time and location context information from the user observed QoS records, the user evaluated records and the user feedback data, so as to construct the user-service-time/location 3-dimension data. Secondly, due to each user only invokes one or several cloud services out of the numerous candidates at each time, resulting in many missing QoS values in history records, the project proposes a tensor decomposition algorithm which is able to deal with the triadic relations of user-service-time/location model. Thirdly, to solve the high sparsity of the available QoS data, the project presents a tensor decomposition model based on the time similarity and location similarity. Lastly, two prediction models are aggregated together to get more accuracy result and the extensive experiments are conducted. Overall the project can help the service users to select best QoS-aware services and help the service providers to recommend services to the potential customers.

英文关键词： QoS Prediction;Time-aware;Location-aware;Cloud Service;Service Recommendation