基于主题网络的用户内在兴趣发现及演进研究

项目名称： 基于主题网络的用户内在兴趣发现及演进研究

项目编号： No.61502247

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： 李华康

作者单位： 南京邮电大学

项目金额： 21万元

中文摘要： 个性化服务作为互联网研究热点之一，主要集中在用户建模、行为分析、兴趣漂移和个性化推荐等方面。这些研究主要采用大量历史用户数据建立用户模型并以此对个体用户进行模式匹配与预测，忽视了用户作为智能体对兴趣具有内在认知、加工、演化的过程。本项目拟通过构建一个面向互联网内容的全局兴趣主题网络，并以此探索用户个体内在兴趣发现及演进的基础理论和关键技术。具体研究包括：1）通过关键字提取技术、分析关键词与兴趣主题之间的关联关系并以此构建兴趣主题间关联，特征提取和机器学习算法判别主题间的前后因果关系等技术研究兴趣主题网络的构建理论与建模方法；2）利用访问页面频度、停留时间等关键特征研究用户兴趣度量建模方法3）采用GE改进算法发现用户兴趣在全局兴趣网络上的内在兴趣，揭示用户兴趣内在的演进和更新规律。该项目的研究成果丰富用户行为分析的研究内涵，对推动个性化服务的应用有着重要的意义。

中文关键词： 个性化服务；用户建模；内在兴趣；主题网络；兴趣演进

英文摘要： The research of personalized service, as one of the most popular researches in Internet era, are focused on user profile, behavioral analysis, interest drift and personalized recommendation and so on. These studies are mainly to individual user’s pattern based on massive historical user data to make a forecasting. They ignore the user as an agent can generate new interests from the current knowledge base..This project intends to build a global interest topic network based on Internet content, which is used to explore the basic theory and key technology of the development and evolution of user’s inherent interests. Specific studies include: 1) in order to study the construction theory and modeling algorithm, keyword extraction techniques, analyzing of association between keywords and interest topic to build inter-related interest topics, feature extraction and machine learning algorithm will be used to determining the causal relationship between interest topics; 2) using page accessing frequency, residence time and other key feature to research interest measurement modeling method; 3) using improved GE algorithm to discover user’s inherent interest in the global interest network, revealing the basic evolution and update pattern for users inherent interest..The achievements of this research project can enrich user behavior analysis research content, and promote the application of personalized services significantly.

英文关键词： personalized service;user profile;inherent interest;topic network;interest evolution