融合用户兴趣漂移特征的个性化推荐研究  

Research on Personalized Recommendation Combining User Interest Drift

作　　者：蒋武轩 易明[1] 汪玲 Jiang Wuxuan

机构地区：[1]华中师范大学信息管理学院,湖北武汉430079

出　　处：《情报理论与实践》2022年第1期38-45,37,共9页Information Studies:Theory & Application

基　　金：国家社会科学基金重点项目“在线健康社区知识共创机理及引导机制研究”研究成果,项目编号:21ATQ006。

摘　　要：[目的/意义]基于社交网络用户短期兴趣和长期兴趣,挖掘用户不同时间窗口下长短期兴趣,能够提高用户兴趣发现的准确性,解决推荐系统不能适应用户兴趣变化的问题。[方法/过程]通过对社交网络用户兴趣的研究发现,社交网络用户兴趣可以分为短期兴趣和长期兴趣,据此构建融合用户兴趣漂移特征的个性化推荐模型。采用时间窗口的方法挖掘用户短期兴趣,利用遗忘曲线跟踪用户长期兴趣变化。在此基础上对用户进行聚类,根据用户聚类结果为用户推荐兴趣相似用户。并以微博真实数据为例进行实证。[结果/结论]融合用户兴趣漂移特征的个性化推荐模型能够较准确地发现用户兴趣漂移特征,满足用户个性化信息需求。[局限]仅使用"微博"这一应用广泛的网络社交平台进行实证,未能从多个网络社交平台进一步验证模型的可行性和准确性。[Purpose/significance] Based on the short-term and long-term interests of social network users, mining the long-term and short-term interests of users under different time Windows can improve the accuracy of user interest discovery and solve the problem that the recommendation system cannot adapt to the change of users’ interests.[Method/process] Through the study of social network users’ interests, it is found that social network users’ interests can be divided into short-term interests and long-term interests, and based on this, a personalized recommendation model integrating user interest drift characteristics is built.The method of time window is used to mine users’ short-term interest and the forgetting curve is used to track users’ long-term interest.On this basis, user clustering is carried out, and users with similar interests are recommended according to the clustering results.And take the real data of Weibo as an example to carry on the demonstration.[Result/conclusion] Personalized recommendation model integrating user interest drift features can find user interest drift features more accurately and meet user’s personalized information needs.[Limitations] This paper only uses “Weibo”,a widely used social network platform, to conduct empirical study, failing to further verify the feasibility and accuracy of the model from multiple social network platforms.

关 键 词：兴趣漂移 行为特征 人类行为动力学 个性化推荐 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]