融合用户偏好和信任关系的混合群组餐厅推荐

Hybrid Group Restaurant Recommendation Based on User Preference and Trust

在推荐系统领域,群组推荐可以有效解决传统个性化推荐存在的仅对单用户推荐、数据稀疏、计算量大等问题,已成为该领域研究的热点。在传统的推荐算法中,仅使用用户的评分数据,没有考虑用户的信任关系。本文提出的算法通过引用用户偏好模型,获取用户-餐厅偏好评分矩阵,代替原始的用户-餐厅评分矩阵,降低了数据的稀疏性;在计算用户信任度时,从公平性、准确性、影响力考虑了不同因素对用户信任度的影响;通过K-means算法对同城的用户进行聚类,使用改进的比重偏好融合策略预测群组偏好进行推荐。通过实验可知,该方法有更好的推荐结果。

In the field of recommended system,group recommendation can effectively solve the problems of traditional personalized recommendation,such as single user recommendation,sparse data,large amount of calculation and so on.In the traditional recommendation algorithm,only the user's rating data is used;however,the user’s trust relationship is not consid-ered.By quoting the user preference model,the user restaurant preference score matrix is obtained instead of the original userrestaurant score matrix,which reduces the sparsity of data;when calculating user trust,we consider the influence of different factors on user trust from fairness,accuracy and influence;the users in the same city are clustered by K-means algorithm,and the group preference is predicted by using the improved proportion preference fusion strategy for recommendation.Ex-periments show that the method has better recommendation results.