面向电子商务协同推荐的新型用户兴趣模型研究

项目名称： 面向电子商务协同推荐的新型用户兴趣模型研究

项目编号： No.71202165

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 工商管理

项目作者： 李聪

作者单位： 四川师范大学

项目金额： 19.5万元

中文摘要： 用户兴趣模型是实现电子商务协同推荐的核心和基础。但传统的评分矩阵模型会导致稀疏性、兴趣漂移及可扩展性问题，新兴的对分网络模型还只能部分解决之。对此，本项目研究拟构建一种新型用户兴趣模型"多源进化网络"，可望在该模型统一框架下综合解决上述问题：（1）采用源自人工智能理论的智能agent方法来设计多源用户兴趣数据采集机制，并通过构造两级多元函数组对所采数据进行信息融合以解决稀疏性；（2）借鉴仿生计算思想，建立包括养分蔓延策略、养分耗散策略的模型自适应进化机制以解决兴趣漂移；（3）基于图论来设计准确、高效的模型相似性计算方法，在此基础上构建模型最近邻搜寻机制以解决可扩展性。本项目的研究成果不仅可以促进协同推荐理论的完善及其在电子商务中的应用、推广与普及，为新一代电子商务个性化推荐服务提供理论模型、算法实现及系统平台，也有助于构建电子商务智能数据挖掘与复杂机器学习系统。

中文关键词： 协同推荐；用户兴趣模型；多源进化网络；电子商务推荐系统；信誉

英文摘要： User interest model is fundamental for implementing collaborative recommendation in E-commence. While traditional rating matrix leads into several problems including sparsity, interest drifting and scalability, the emerging model (bipartite network model) can only resolves small portion of the above problems. Hence, we will construct a novel user interest model called "Multisource Evolving Network" (MEN) to solve the sparsity, interest drifting and scalability problems comprehensively: (1) By using intelligent agent method, originating from artificial intelligent theory, the mechanism of multisource user interest data will be designed; while an information fusion method based on the set of two-stage multivariate function is to be built to solve sparsity problem for obtained data; (2) By lessons from bio-inspired computing, we will build the adaptive evolving mechanism of MEN, which including nutrient spread tactic and nutrient dissipation tactic, to solve interest drifting problem; (3) By using graph theory, we can design an similarity computing method with high accuracy and efficiency for MEN, and then we will build the neighborhood searching mechanism of MEN to solve scalability problem. The results can be used to consummate and popularize the application of collaborative recommendation, while providing novel

英文关键词： collaborative recommendation；user interest model；multisource evolving network；E-commerce recommender systems；reputation