云环境用户多兴趣图谱的移动商务关联性推荐模型及算法研究

项目名称： 云环境用户多兴趣图谱的移动商务关联性推荐模型及算法研究

项目编号： No.71271186

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 管理科学

项目作者： 张亚明

作者单位： 燕山大学

项目金额： 50万元

中文摘要： 移动互联网爆炸式增长、电子商务迅猛发展以及智能手机的快速普及，使"移动互联新生态"在全球范围内迅速成长，其个性化推荐系统也跻身为提高企业竞争力、满足用户即时个性化需求的利器。移动商务的特殊性使传统推荐系统难以简单移植并满足"数字宇宙"时代的特殊需求。课题基于云环境下用户多兴趣图谱和"关联性"构建移动商务推荐模式和算法。首先，从用户情景关联度、语义关联度提出面向兴趣的多源信息空间模型，深入挖掘云环境下移动用户兴趣漂移与演变规律；其次，基于信息熵方法构建移动用户信息熵矩阵与多兴趣参量矩阵，提出兴趣细粒度特征的提取及表述方法，破解复杂网络环境下移动用户多源兴趣特征提取、关联信息传递及多兴趣测度的动态映射难题；最后，采用分布式处理框架MapReduce设计移动商务多兴趣关联的并行推荐算法，解决并突破"关联时代"传统推荐系统弊病，以分布式形式为用户提供信息粒度更为精细、情景更为关联的便捷式推荐服务。

中文关键词： 移动商务；云环境；关联性推荐模型；多兴趣图谱；推荐算法

英文摘要： With the explosive growth of mobile Internet, rapid development of electronic business and rapid popularization of intelligent mobile phone, a mobile Internet new ecology is emerging globally. Therefore, its personalized recommender system is a good tool to improve operators' competitiveness and meet users' personalized needs. However, the particularity of mobile commerce makes it difficult to simply transplant traditional recommendation system to mobile commerce and meet the special needs in Digital Universe era. This project proposed a recommendation model and related algorithms based on user multi-interest map in cloud environment. Firstly, a multi-source information space model was proposed from the aspects of user scenario and semantic correlation degree, thus deeply analyzing the interest shift and evolution of mobile users. Secondly, a mobile user information entropy matrix and the interest parameter matrix were proposed based on information entropy method to breakthrough the feature extraction, information transfer and dynamic interest mapping problem in complicated network environment. Finally, a distributed processing framework Map Reduce was implemented to design the parallel recommendation algorithm for mobile commerce to breakthrough the bottlenecks of traditional recommendation systems, and provide

英文关键词： Mobile commerce；Cloud environment；Association recommendation model；Muti- interest map；Recommendation algorithm