融合深度神经网络和张量分解的地点推荐算法  

Location recommendation algorithm combined with tensor factorization and deep neural network

作　　者：肖述 张䶮 周文荣 XIAO Shu;ZHANG Yan;ZHOU Wen-rong(School of Computer Science and Information Engineering,Hubei University,Wuhan 430062,China;Department of Information Development and Management,Hubei University,Wuhan 430062,China)

机构地区：[1]湖北大学计算机与信息工程学院,湖北武汉430062 [2]湖北大学信息化建设与管理处,湖北武汉430062

出　　处：《计算机工程与设计》2022年第1期171-178,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61977021);湖北省2019年技术创新专项基金项目(2019ACA144)。

摘　　要：为提高利用张量分解技术进行基于位置社交网络的地点推荐的推荐性能,提出一种利用张量分解技术且融合神经网络的地点推荐算法。融合多层感知机和长短期记忆网络基于张量分解技术建模用户的签到行为,将学习到的用户偏好表示馈送到推荐生成器和推荐判别器组成的对抗生成网络中,通过对抗训练学习最佳用户偏好表示用于推荐。基于真实数据集的实验验证了该算法的有效性和高效性。To solve the problem of low performance of location recommendation in location-based social network using tensor factorization technology,a location recommendation algorithm combining neural network and tensor factorization was proposed.The multi-layer perceptron and long-and short-term memory network were combined to model the user’s check-in behavior based on tensor factorization technology.The learned user preference representation was fed into a generated countermeasure network composed of a recommendation generator and a recommendation discriminator,and the optimal user preference representation was learned through confrontation training for recommendation.Experimental results based on real data sets verify that the proposed recommendation framework is superior to the state-of-the-art location recommendation model.And the accuracy of the algorithm is improved.

关 键 词：位置社交网络 张量分解 长短期记忆网络 对抗生成网络 多层感知器 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]