摘要： 针对推荐系统中用户的个性化需求,提出一种基于用户兴趣三维建模的个性化推荐算法。通过分析用户行为数据,从兴趣广度、兴趣深度和兴趣时效3个角度分析用户的兴趣构成,对用户兴趣进行三维建模,并在此基础上,逐步添加维度,设计用户之间兴趣相似度的三级计算方法。在真实推荐系统数据集上的实验结果表明,用户兴趣三维模型比一维模型、二维模型更能准确地表征用户兴趣,基于用户兴趣三维建模的个性化推荐算法能够提高个性化推荐的准确率。

关键词: 个性化推荐, 用户兴趣三维建模, 兴趣广度, 兴趣深度, 兴趣时效, 用户兴趣相似度

Abstract: To realize the personalized recommendation,an algorithm based on three-dimensional user interest modeling is presented.First,by analyzing user’s behavior data,three aspects of user interest are analyzed for building three-dimensional user interest model,including interest width,interest depth and user interest timeliness.Secondly,based on the three-dimensional user interest model,the dimension is gradually added and a method for calculating the interest similarity between two users is proposed.At last,a personalized recommendation algorithm is given.Experiment uses the data set derived from real recommendation system.The results show that,three-dimensional user interest model can describe user interest more accurately than two-dimensional and one-dimensional user interest model,and the proposed personalized recommendation algorithm can improve the accuracy rate of personalized recommendation.

Key words: personalized recommendation, three-dimensional user interest modeling, interest width, interest depth, interest timeliness, user interest similarity

中图分类号: 

• TP312