【作者】 沈杰峰；

【导师】 杜亚军；

【作者基本信息】 西华大学 ， 计算机软件与理论， 2006， 硕士

【摘要】 Internet技术的发展为人们提供了方便快捷的信息获取手段。面对网络上的海量信息,人们有时往往会感到无所适从。如何为每个用户提供快捷准确,满足个人实际需要的信息,己成为众多业内人士越来越关注的问题。 基于信息过滤技术的信息推荐系统在与用户交互时,针对不同的用户采取不同的服务策略,提供不同的服务内容。在信息领域,每个用户都有自己特定的、长期的信息需求,用这些信息需求组成过滤条件,从动态的信息资源流中过滤出服务需求的内容,屏蔽掉无用的信息并进行服务。 近年来推荐系统在理论和实践中都得到了很大发展。但是随着系统规模的扩大,信息推荐系统也面临一系列挑战。针对这些挑战,本文进行了有益的探索和研究,研究的主要内容为如下两方面: 1).在推荐系统推荐质量研究方面,提出了一种基于项目特征评价的协同过滤算法,通过项目特征分解的方法来提高系统在评价数据稀疏情况下的系统性能,并通过用户聚类的方法来减少搜索空间,减少在线计算的复杂度。实验证明,本算法可以有效改善推荐系统的推荐精度,并在一定程度上改善系统的反映时间。 2).设计了一种基于蚁群算法的网页推荐算法,它把基于内容的信息过滤和协同过滤两者的优点结合在了一起,弥补了各自的缺陷。实验证明,如果在协作式过滤方法中融入基于内容过滤的技术,系统性能将有所提高。通过两者的结合,可以获得基于内容过滤的优点,包括能进行覆盖所有文档和用户的早期预测;同时也能获得协作式过滤的优点,即使用户评估过的文章数不断增加,系统仍能给出精确的预测。更多还原

【Abstract】 With rapid development and popularity of the Internet, It is more and more convenience to find the needed information for everyone. The Internet information is so great that people often lost themselves in it. Therefore how to provide the personal information to people is becoming the important problem that the reaseachers care for.The information recommender system based on information filtering technology communicates with the users, and provide different information for different users .In information field, every user have his steady perennial requirement. The requirement of users can contribute to filter the useless information and help users to select the needed ones.Recommendation systems have been very successful in both research and practice, but some challenging problems remain in this field. Aimed at the main challenges of recommendation systems, this thesis explored some key technologies of recommendation systems. The main research works in the thesis cover two aspects:1). For the improvement of the recommendation quality, we proposed a collaborative filtering recommendation algoritm based on the rating of items’ character. In the algoritm, items are parsed to characters, and as contributes to improve the system performance when rating data is sparcity. In addition, the cluster of the users is used to minish the searching range for the active user. The experiment results suggested that this method could efficiently overcome the extreme sparcity of user rating data and provide better recommendation results than traditionalcollaborative filtering algoritms2). For making use of the advantages of the collaborative-filtering algoritms and content-based filtering algoritms, we proposed a multi-filtering recommendation algoritm using ant colonies. The experiment results suggested that the system’s performance is improved after the combine of collaborative-filtering and content-based algoritm. The algoritm contain the advantages of content-based algoritm (covering all of documents and and the rating of early users). At the same time, the recommendation is still accuracy even if the increase of the user rating, as is also the advantage of the collaborative-filtering algoritm.更多还原