计算机工程与应用 ›› 2021, Vol. 57 ›› Issue (23): 113-121.DOI: 10.3778/j.issn.1002-8331.2011-0350

• 大数据与云计算 • 上一篇    下一篇

基于相似用户好奇心的多样性推荐方法

田维安,陈红梅,周丽华   

  1. 云南大学 信息学院,昆明 650504
  • 出版日期:2021-12-01 发布日期:2021-12-02

Diversified Recommendation Method Based on Similar Users’Curiosity

TIAN Wei’an, CHEN Hongmei, ZHOU Lihua   

  1. School of Information Science and Engineering, Yunnan University, Kunming 650504, China
  • Online:2021-12-01 Published:2021-12-02

摘要:

推荐技术已经成为信息过载时代提供个性化服务的关键技术。由于推荐结果的多样性可以提升推荐效果,多样性推荐方法开始备受关注。针对现有基于朋友好奇心的多样性推荐方法中,诸如朋友、信任关系等难以获取及比较稀疏的问题,提出了基于相似用户好奇心的多样性推荐方法(SUC)。分析用户的真实评分,计算相似用户集;采用协同过滤方法,计算用户的预测评分;分析用户的真实评分和预测评分,计算用户的好奇心评分;融合预测评分和好奇心评分,计算用户的项目推荐列表。SUC方法不需要额外的用户关系信息,更具普适性。在五个真实数据集上的实验表明,与基准方法相比,SUC方法不仅提高了推荐多样性,同时也提升了推荐准确率、召回率和覆盖率。

关键词: 推荐系统, 协同过滤, 相似度, 好奇心, 多样性

Abstract:

Recommendation technology has become the key technology to provide personalized services in the era of information overload. Since the diversity of recommendation results can improve the recommendation effect, the diversified recommendation method has attracted researcher’s attention. It is difficult to obtain the relationships between users, such as friends and trust, which is used in the existing method based on the curiosity of friends. So, this paper proposes a diversified recommendation method based on Similar Users’ Curiosity(SUC). First, it analyzes the users’ real ratings and calculates the set of similar users. Second, it calculates the users’ predicted ratings based on the collaborative filtering method. Then, it calculates the users’ curiosity ratings by analyzing the users’ real ratings and predicted ratings. Finally, it integrates the predicted ratings and curiosity ratings to calculate the users’ item recommendation lists. The proposed method is more useful because it does not require additional information. Experiments on five real data sets show that compared with the benchmark methods, the SUC method not only improves the diversity of recommendation, but also improves the accuracy, recall and coverage of recommendation.

Key words: recommendation system, collaborative filtering, similarity, curiosity, diversity