Computer Engineering and Applications ›› 2022, Vol. 58 ›› Issue (9): 98-106.DOI: 10.3778/j.issn.1002-8331.2107-0011

• Big Data and Cloud Computing • Previous Articles     Next Articles

Group Recommendation Method Combining Leader Influence and Implicit Trust Metrics

WANG Yonggui, LIN Jiamin, HE Jiayu   

  1. 1.College of Electronics and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China
    2.College of Software, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Online:2022-05-01 Published:2022-05-01

融合领导者影响与隐式信任度的群组推荐方法

王永贵,林佳敏,何佳玉   

  1. 1.辽宁工程技术大学 电子与信息工程学院,辽宁 葫芦岛 125105
    2.辽宁工程技术大学 软件学院,辽宁 葫芦岛 125105

Abstract: Group recommendation needs to consider the preferences of all members of a group at the same time, and then integrate the preferences to recommend items to the group. Most of the existing researches on group recommendation methods use fixed and symmetric relationship weights to predict scores, ignoring the complex relationship influence among group members, which will lead to low accuracy of group recommendation. In order to solve the above problems, a group recommendation method(GRS-IT) is proposed, which integrates leader influence and implicit trust. By combining fuzzy C-means clustering with Pearson correlation, the method can find groups with high similarity, which can effectively improve the effect and stability of group recommendation. By introducing the method of leader influence, combining Pearson correlation and an implicit trust calculation, the leaders in the group are found out and the dynamic influence weights between leaders and members and between members are obtained to reduce the error rate of group recommendation. In addition, this method integrates the time function based on the human forgetting curve into the prediction of item rating, and gives different time weight values over time to the prediction score, which further improves the accuracy of group recommendation. Finally, a comparative experiment is used to verify the effectiveness of GRS-IT. The results show that, compared with other group recommendation methods on the Movielens100K data set, the recommendation results are significantly improved in terms of accuracy and group member satisfaction.

Key words: group recommendation, fuzzy C-means algorithm, leader influence, implicit trust, time function

摘要: 群组推荐需要同时考虑一组内所有成员的偏好,融合偏好进而向群组推荐项目。现有关于群组推荐方法的研究中大多使用固定的、对称的关系权重进行预测评分,忽略了群体成员之间复杂的关系影响,这会导致组推荐准确度偏低。为了解决上述问题,提出了一种融合领导者影响与隐式信任度的群组推荐方法(GRS-IT),该方法通过模糊C均值聚类与皮尔逊相关性结合的方法,发现高相似度群组,有效地提高群组推荐效果和稳定性;引入领导者影响的方法,结合皮尔逊相关性与一种隐式信任度计算找出组内领导者并获取领导者与成员、成员彼此之间的动态影响权重,降低群组推荐的误差率;此外,该方法将基于人类遗忘曲线的时间函数融入到项目评分预测中,预测评分随时间变化赋予不同的时间权重值,进一步提高了群组推荐的准确性。采用对比实验对GRS-IT的有效性进行验证,结果表明,在MovieLens100K数据集上与其他群组推荐方法相比,推荐结果在准确度和群组成员满意度方面都有显著提高。

关键词: 群组推荐, 模糊C均值算法, 领导者影响, 隐式信任度, 时间函数