软件服务化背景下的云服务推荐方法研究

With the fast and wide adoption of cloud computing, many small and medium enterprises (SMEs) and individual users prefer to apply cloud services to build their business system or personal applications. Cloud services have many advantages such as on-demand access, remote accessibility, easy expansion and upgrade, and reduced maintenance and management costs. The recent years have seen a tremendous increase of the type and the number of different cloud services in the Software as a Service (SaaS) environment. However, given the lack of cloud computing technology of SMEs and individual users, it is tedious when manually selecting appropriate services from the large cloud service pool, which can not only achieve the functional requirements but also meet the user's hidden interests. This jeopardizes the future development of cloud services in the SaaS environment..In this project, we address this challenge through automatic user interest discovery and custom cloud service recommendation in the SaaS environment.Some key issues in cloud service recommendation research,such as user's interest modeling, personalized recommendation method and comprehensive evaluation of the cloud recommendation algorithms,will be systematically studied..Firstly, we will establish an analytical model to describe user's interests. Based on the solid ontology modeling theory, the proposed model can capture cloud service user's interest from difference sources, such as transaction information, online behavior and serivice reviews. On the basis, the cloud service user's interest multi-granularity knowledge discovery method based on multi-sources user data will be researched..Secondly, we will study the process of cloud service recommendation based on user's interest. Driven by our our analytic model, we will develop personalized recommendation algorithms for single and combined cloud service application situations in the SaaS environment.The novel algorithms will contribute not only theoretically but also practically to accelerating the wider adoption of SaaS and Cloud Computing..Thirdly, we will develop a evaluation metric to assess the effectiveness of different cloud service recommendation algorithms, and then propose a multi-attribute evaluation method. The new multi-attribute evaluation method intergrates both the user's interest and algorithm performance in the SaaS environment, which enables us to conduct comprehensive evaluation of existing and new cloud service recommendation algorithms..Based on the results of our theoretical analyses, we will also design and prototype cloud service recommendation systems for SMEs and individual users.In summary, this project not only advances the systematic analysis of cloud service recommendation, but also significantly expand the scope of Service Science, Cloud Computing and Recommendation System.

云计算理论与技术的快速发展，使得越来越多的中小企业和个人用户开始选择按需定制、配置灵活、动态可扩展且更新与维护代价低廉的各类云服务。而对于专业知识相对匮乏的用户而言，快速选择既可达到特定功能需求又能满足隐性用户兴趣的云服务是极其困难的，迫切需要研究相关的云服务推荐问题。本项目针对云服务应用的新特点，从云服务用户兴趣建模、个性化推荐方法以及推荐算法综合评价等方面，系统研究软件服务化背景下的云服务推荐方法。主要研究内容包括：建立云服务的用户兴趣表示模型，研究基于交易信息、在线行为和服务评论等多源数据的用户兴趣多粒度知识发现方法；探索软件服务化背景下的云服务推荐过程模型，研究设计面向单一及组合应用的云服务个性化推荐算法；分析传统推荐算法度量指标的适用性，并建立考虑用户兴趣满足的推荐算法新度量，进而研究云服务推荐算法的综合评价方法，为软件服务化背景下云服务的用户兴趣发现与个性化推荐提供理论支持。

云计算技术的日趋成熟以及智能终端设备、移动互联网应用的迅猛发展，使得云服务已经成为人们工作生活中的重要技术载体和推动力，云计算环境下面向服务的开放式创新也正在掀起新一轮产业变革。云服务智能优化推荐是解决云服务选择过程中信息超载问题的重要途径，也是开放平台运营商进行自动化大规模定制的关键路径。然而，受服务数据规模、交易规范性、平台成熟度等方面的影响，现有云服务推荐系统及方法的实际效果并不理想，且通常具有较强的应用环境约束。.本项目重点围绕软件服务化背景下开放平台中的云服务个性化选择与动态推荐问题，集成应用QoS分析与协同过滤推荐、信任建模与动态演化、排序预测与时间序列预测、复杂网络建模与博弈分析等方法，系统研究了多目标优化推荐、合作演化博弈、信任网络分析、大数据聚类、多目标匹配决策等多项支持云服务选择与推荐的基础性理论方法，重点围绕软件服务化背景下开放平台中的云服务个性化选择与动态推荐问题，研究构建了考虑多维QoS性能的云服务匹配与推荐系统、集QoS预测和客户满意度估计的云服务可信性评估模型、面向云服务选择的多属性可信评估决策支持方法、信任增强型云服务选择优化模型等多项创新性理论成果，并依托WS-DREAM开放式云服务QoS性能数据测试集等，展开了十分丰富的模型及算法验证工作。.通过近三年的前沿理论研究与技术攻关工作，本项目组已经在《Knowledge-based Systems》、《Applied Mathematics and Computation》、《Chaos, Solitons & Fractals》、《IEEE Transactions on Fuzzy Systems》、《PLOS One》、《Physica A》等国际期刊和《系统工程理论与实践》、《中国管理科学》等国内期刊发表相关学术论文16篇，其中：SCI收录13篇，国家自然科学基金委管理学部认定的A类期刊2篇；申请相关的国家发明专利4项，其中：已授权2项，已进入实审阶段2项；参加国内外学术会议10人次，在中国工程院学部学术活动上做大会报告1次；培养青年骨干教师3人、博士研究生7人、硕士研究生7人。.通过本项目的研究，为解决软件服务化云服务用户兴趣发现与个性化推荐问题提供新的方式和途径，也有助于丰富和深化云服务推荐系统的理论研究体系，具有重要的理论意义和实际价值。