基于社交网络的用户需求发现与物品推荐
2016-11-14林梦迪
林梦迪



摘要:微博作为当下最受欢迎的社交网络之一,包含了大量的用户需求和兴趣偏好信息,如何动态地从微博内容中提取用户的需求和偏好信息,将推荐算法结合社交网络产生推荐结果,解决信息过载的问题,目前暂时还没有相关的较为成熟的应用。本文设计并实现了基于社交网络的物品推荐系统,提取用户微博内容关键词作为用户需求特征,建立物品信息库,通过文本相似度计算用户需求和物品信息之间的匹配度,采用基于内容的推荐算法产生推荐结果。最后进行离线实验,对推荐系统产生的推荐结果进行评测分析。
关键词:社交网络;用户需求;基于物品推荐算法;微博
中图分类号:TP393 文献标识码:A 文章编号:1009-3044(2016)24-0260-03
Abstract: Microblog is one of the most popular social networks, containing plenty of information of users preference and needs. However, there are still no mature applications to extract the information of users preference and needs from microblogs and combine those with recommendation algorithms to recommend items. Thus, this paper proposes and implements a recommendation system for social networks. First of all, users content of weibo was crawled and keywords were extracted as characteristics of users needs. Next, an item repository was built with specific characteristics of the items. Using the text similarity algorithm, similarity between users needs and items characteristics can be computed. Then with the content-based recommendation algorithm, we produced recommendation results for users which they may be interested in. At last four offline experiments on the recommendation results were done to evaluate and analyze the performance of this recommendation system.
Key words:social networks; users interests; content-based recommendation algorithm; Microblog
1 背景
近年来,社交网络的发展引人注目。目前据百度百科的统计,约有一半以上的中国网民通过社交网络进行沟通交流、分享信息。据尼尔森2010年的报告,全球用户在互联网上22%的时间花费在社交网站和社交媒体上[1]。通过社交网络不仅可以很好地获取用户社交关系,并且允许用户公开地展现自己,表达用户个人兴趣,对于研究个性化推荐系统具有很好的价值。本文利用社交网络信息对用户进行个性化物品推荐。
目前流行的推荐算法主要有基于内容的推荐算法[2]、协同过滤推荐算法[3]和混合推荐算法等。本文采用基于内容的推荐算法,其不需要获取大量的用户评分数据,因此不存在评分数据稀疏性问题;……
