【摘要】 推荐系统在互联网技术快速发展的今天变得越来越重要,因为它能为用户做出最适当的选择。协作过滤(CF)是推荐系统设计中应用最成功和最广泛的技术,能根据用户的过去评级记录,推荐活跃用户的项目。不过,与用户项目矩阵中的大量用户和项目相比,用户对项目的评级非常稀疏,CF可能会导致不佳的建议。它将用户的隐式交互记录与项目相结合,通过采用关联规则来挖掘处理大量数据,可以捕获每个交易的多个采购关联规则,而不仅仅是计算总的采购量,并通过实施修改了的预处理,在基于多次购买完成的用户之间发现类似的兴趣模式。另外,随着关联规则挖掘的表现,聚类技术已被用于减少数据的大小和项目空间的维度,然后,计算出基于其特征的项目之间的相似性,以提出建议并进行实验。结果表明,在Precision和Recall指标两个方面,即使数据非常稀疏,这种技术也能达到很好的性能。更多还原

【Abstract】 The recommendation system now becomes more and more important in the rapid development of Internet technology because it can make the most appropriate choice for users. CF(collaborative filtering)is the most successful and widely-used technique in recommended system design, and it can recommend active users’ projects on the basis of their past rating records. However, compared to the large number of users and projects in the user’s project matrix, the user’s ratings for projects are very sparse, and CF may lead to poor recommendations. It combines the user’s implicit interaction record with the project, and uses association rules to mine and process large amounts of data, multiple purchasing association rules for each transaction can be captured, rather than just calculating the total procurement volume. By implementing the modified pre-processing,similar patterns of interest are found among users based on multiple purchase completions. In addition, with the performance of association rule mining, clustering technology is used to reduce the size of data and the dimensions of project space. Then, the similarity among the items based on their characteristics is calculated, so as to make recommendations and conduct experiments. The experiment results indicate that in the two aspects of Precision and Recall indicators, even if the data is very sparse, this technology can achieve good performance.更多还原