PS-LSTM:Popularity Analysis And Social Network For Point-Of-Interest Recommendation In Previously Unvisited Locations
Chongyu Zhong, Jinghua Zhu, Heran Xi · 2021
With the development of technology and the gradual convenience of transportation, due to various reasons such as travel or business trip, people often leave their familiar areas and visit relatively unfamiliar areas in real life.It will be presents a new problem to the recommendation algorithm, namely the POI recommendation in previously unvisited locations. In order to make the recommendation system still provide users with high quality POI recommendation in this situation, this paper proposes a POI recommendation algorithm based on deep learning and popularity analysis.In the model, we integrate the popularity of POI and social network to alleviate the cold start problem .When integrating social information, we use attention mechanism to measure the influence of different friends on users, and then weigh it to get the social impact. When fusing the information of POI popularity, we consider that the influence of POI popularity varies with geographical distance and time, so we construct a dynamic popularity influence vector in the model.For making better utilize of the characteristics of the check-in sequence, we use the recurrent neural network to model the spatiotemporal sequence of the users’ POI check-in. Finally, MLP is used to integrate social impact, popularity characteristics and sequence information to recommend for users.The proposed algorithm is verified on yelp which is a real city dataset and compared with several classic POI recommendation algorithms.Experimental results show that the proposed algorithm could achieve a better accuracy.