一种融合地理位置信息的协同过滤推荐算法

发布时间：2018-02-22 11:51

本文关键词： 推荐系统协同过滤地理位置信息邻居模型隐参数模型　出处：《中文信息学报》2016年02期 　论文类型：期刊论文

【摘要】：目前,基于用户消费数据构建的推荐系统在电子商务领域发挥着越来越大的作用,而在这些数据中,商家本身具有的地理位置信息忠实地记录了用户的消费痕迹,能够有效反映出用户在地理位置维度上的个人偏好信息,从而对推荐系统具有非常重要的意义。现有工作一般只利用了用户对地点的评价以及地点之间的距离,无法反映出不同地点之间的关联关系,以及用户在不同地点中的偏好权重问题。该文从地理区域划分的角度出发,研究了用户在区域范围内的消费兴趣偏好,以及不同粒度级别的区域划分方法对推荐模型的影响,探索了在推荐过程中有效融合地域信息的方法,考虑了包括地区的全局性影响、用户对地区的偏好等,结合这些因素提出了融合地理位置信息的推荐模型LGE、LGN及LRSVD。通过在Yelp数据集上的实验表明,这些模型相比于传统的推荐算法能够有效提高预测效果。
[Abstract]:At present, the recommendation system based on user consumption data is playing a more and more important role in the field of electronic commerce, and in these data, the geographical location information of the merchant itself has faithfully recorded the user's consumption trace. It can effectively reflect the user's personal preference information in the geographical dimension, which is of great significance to the recommendation system. The existing work generally only uses the user's evaluation of the location and the distance between the locations. The relationship between different locations and the preference weight of users in different locations can not be reflected. This paper studies the preference of consumer interest in different geographical areas from the point of view of geographical region division. As well as the effect of different granularity on the recommendation model, this paper explores the method of effectively merging the regional information in the process of recommendation, including the overall impact of the region, the user's preference for the region, etc. Based on these factors, a recommendation model LGELGN and LRSVD are proposed. The experiments on the Yelp dataset show that these models can effectively improve the prediction effect compared with the traditional recommendation algorithm.
【作者单位】：国家计算机网络与信息安全管理中心;中国科学院大学;中国科学院信息工程研究所;University
【基金】：国家自然科学基金青年基金(61402466)
【分类号】：TP391.3

【相似文献】