移动计算环境下基于动态上下文的个性化Mashup服务推荐  

A new paradigm for personalized mashup recommendation based on dynamic contexts in mobile computing environments

作　　者：何伟[1,2] 崔立真[1,2] 任国珍[1] 李庆忠[1,2] 李婷[1] 

机构地区：[1]山东大学计算机科学与技术学院,济南250101 [2]山东大学山东省软件工程重点实验室,济南250101

出　　处：《中国科学：信息科学》2016年第6期677-697,共21页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:61572295;61303085);科技部创新方法专项(批准号:SQ2015IMC600006);山东省自然科学基金(批准号:ZR2014FM031;ZR2013FQ014)资助项目

摘　　要：移动互联网的快速发展和智能终端数量的迅速增长使得Internet服务更加密切地融入到人们的生活中,面向非专业用户的Mashup服务在移动计算环境下有着更广泛、迫切的需求,同时,用户行为过程中的上下文动态不确定性、用户行为模式复杂性和个性化特征等因素对传统Mashup技术提出了挑战.针对目前Mashup技术和研究中存在的问题,本文将关注点从离散的服务推荐扩展到Mashup构造与运行过程,提出一种新的基于迭代式自主构造的Mashup服务组合和运行模式,主要思想是对大量离散、孤立的历史执行轨迹进行挖掘处理,建立上下文相关和偏好相关的用户行为选择概率模型,为构造优化的个性化Mashup组合方案提供支持.分析和模拟实验表明,面对移动环境上下文动态性和普适性特点,本文方法能够有效简化Mashup构造过程中用户参与的复杂性,减少对用户专业知识的依赖;同时,在用户行为模式预测和Mashup服务构造与推荐方面能够获得比其他服务推荐或组合方法更好的效果.With the rapid development of mobile Internet and increasing number of intelligent terminals, Internet services are now integrated into people＇s daily lives. In mobile computing environments, challenges have emerged for unprofessional user oriented mashup servers with new features, such as dynamic contexts and user preference. In this paper, we focus on the entire construction process rather than individual service recommendation for mashups, and propose a context-based and user preference-based autonomous construction approach for mashups. The main idea is to construct a context- and preference-related conditional probability model for user behavior patterns based on pattern mining of historical mashup logs, which can provide support for the generation of optimal personalized mashup solutions and also recommend suitable web components. Analysis and simulation experiments conducted indicate that the proposed approach can autonomously generate personalized mashups with quality components without dependence on user professionalism.

关 键 词：移动计算环境 上下文 MASHUP 服务推荐 协同过滤 

分 类 号：TP393.09[自动化与计算机技术—计算机应用技术]