具有商品组合效应的集合推荐问题及关键技术研究

项目名称： 具有商品组合效应的集合推荐问题及关键技术研究

项目编号： No.61300137

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 蔡毅

作者单位： 华南理工大学

项目金额： 23万元

中文摘要： 随着互联网技术和电子商务的发展，推荐系统在其中变得越来重要。目前推荐系统的研究都集中在如何向单个用户推荐单个他最喜欢或最可能购买的商品。然而，在现实中单个用户在购买或消费商品的时候往往会有购买一组商品（一个商品集合）的需求，我们把向一个用户推荐一组相关商品的这类推荐称为商品集合推荐。集合推荐和目前推荐系统领域主要研究的单个商品推荐很不同，多个商品的组合含有商品组合效益。即若干个用户都喜欢的单个商品，组合一起成为商品集合打包向用户推荐的时候，用户未必会喜欢，甚至会厌恶。在本研究项目中，我们拟对商品集合推荐问题进行建模和形式化定义；基于经济学中著名的投资组合理论来设计一个候选推荐集合筛选算法，来从所有商品中挖掘和选择出合适的待推荐商品集合；此外，拟研究商品组合效益问题及其计算方法，量化其对集合推荐的影响；提出一个创新的、基于投资组合理论和商品组合效应的推荐算法来给用户进行集合推荐。

中文关键词： 集合推荐；组合效应；投资组合理论；用户建模；

英文摘要： With the development of Internet technology and e-commerce, recommender system becomes more and more important. Existing recommendation methods focus on recommending single item to a user. However, there are many cases that we need to consume a set of items at one time in our real life. For example, many restaurants may provide some food set, and we can choose different salad, main course, dessert, soup and so on. One common distinct feature of these cases is that what we want to recommend to a user is a set of items (e.g., dishes and interest places). In other words, there are many cases in which users consume a set of goods instead of only a single good at a time. We consider that recommending a set of items at a time to a user is a new problem which is different from traditional single item recommendation and we name it as Set Recommendation. In Set Recommendation, there is a group effect while combining multiple items to a set.In other words, a user may like some items individually, however, while combinig them into a set and recommending the set to the user, the user may not like the set of items, even dislike the set.In this research work, we plan to model the problem of Set Recommendation and formally define the problem. Based on the well-known portfolio theory in economics, we propose an algorithm to sel

英文关键词： Set Recommendation；Group Effect；Portfolio Theory；User Profiling；