融合用户行为网络信息的个性化餐馆推荐  

Personalized restaurant recommendation combining user behavior network information

作　　者：傅晨波[1] 郑永立 周鸣鸣 宣琦[1] FU Chenbo;ZHENG Yongli;ZHOU Mingming;XUAN Qi(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023

出　　处：《浙江工业大学学报》2020年第5期574-580,共7页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(11505153,61572439);浙江省自然科学基金资助项目(LR19F030001)。

摘　　要：目前主流推荐算法是在显式评分或者隐式特征相似的基础上,忽略了用户就餐行为序列中的网络拓扑结构关系,使得用户就餐行为中隐含的喜好信息难以被很好地定性刻画。对此,提出了一种融合用户行为网络信息的个性化餐馆推荐系统。首先根据用户的历史就餐行为序列信息,构建其就餐地理位置转移网络和口味信息转移网络。然后利用网络图表征方法得出位置和口味标签的向量表征,刻画了用户历史就餐行为偏好,并提出用户每次就餐时的怀旧指数GT。最后,将怀旧指数GT结合用户评分信息融入到已有的协同过滤推荐算法框架中,得到改进的融合用户行为网络信息的个性化餐馆推荐模型。在Yelp数据上的实验证明:该模型的推荐效果高于传统的基于评分的推荐算法和最好的图嵌入推荐算法。At present,the mainstream recommendation algorithms are based on explicit rating or implicit similarity,the network topology relationships in the sequence of user dining behavior are ignored.It is difficult to describe the implicit preference information in user dining behavior qualitatively.To solve this problem,a personalized restaurant recommendation system combining the user behavior network information is proposed.Firstly,according to the user’s historical dinign behavior sequence information,the dining geographic location migration network and taste information migration network are constructed.Then,the vector representation of the location and taste tag can be obtained by using the network graph representation method,and the historical dining behavior preference of user can be depicted,and the nostalgia index GT of user at each dining is proposed.Finally,the nostalgia index GT combined with user rating information is integrated into the existing collaborative filtering recommendation algorithm framework and an improved personalized restaurant recommendation model integrating user behavior network information is obtained.The experiments on Yelp data prove that the recommendation effect of this model is higher than the traditional recommendation algorithm based on rating and the best recommendation algorithm based on graph embedding.

关 键 词：餐馆推荐 网络构建 网络嵌入 协同过滤 Yelp 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]