融合多源异构数据的混合推荐模型

A Hybrid Recommendation Model Based on Fusion of Multi-Source Heterogeneous Data

为更全面地反映用户个人偏好,提高推荐的准确度,提出了一种融合多源异构数据的混合推荐模型.综合考虑了用户社交关系和用户评论对用户评分的影响,从评论中提取主题信息作为用户和商家的特征,采用社区发现算法为用户划分社区,利用机器学习方法为社区建立模型,预测用户对商家的评分,再根据评分对商家进行排序,取前N个商家推荐给用户.实验结果表明,提出的混合推荐模型与经典推荐算法相比,可提高评分预测的准确度,从而提高推荐的准确度.

In order to reflect users’personalized preferences more comprehensively and improve the accuracy of recommendation,a hybrid recommendation model based on fusion of multi-source heterogeneous data is proposed.This model takes the impacts from both users’social relationships and reviews on ratings into account.Topics are extracted from reviews as user features and business features,and then communities are divided for users via a community discovery algorithm.Finally,a machine learning algorithm is used to model user communities in order to predict ratings.Businesses are ranked based on predicted ratings and then the top N businesses are recommended to the user.The experimental results show that the proposed hybrid recommendation model can improve the rating prediction accuracy and recommendation accuracy compared with the conventional recommendation algorithms.