Research On The Next POI Recommendation Based On Deep Neural Network

With the popularity and rapid development of global positioning system technology,smart mobile devices,and Web 2.0 technologies,location-based social services have gradu-ally become an indispensable tool for daily social life and traveling.In order to extract and filter out information that is of interest to users and meet individualized needs from the mas-sive information flow,point-of-interest recommendation has become an emerging research hotspot.Compared to traditional recommender system,the data of next point-of-interest recommendation is more sparse.At the same time,the user's check-in sequence mode is difficult to characterize and is affected by a variety of contextual factors.Although some work has been proposed,there is no satisfactory method yet.Therefore,this paper further studies the recommendation method for next point-of-interest recommendation based on the existing research work.This paper uses the gating mechanism to model the time and distance characteristics of the user historical check-in sequence;combine with the context information to alleviate the data sparsity problem to improve the recommendation performance of the next point-of-interest recommendation.The main work of this paper is as follows:(1)Analyze the background and meaning of point-of-interest recommendation and the next point-of-interest recommendation,as well as the research status at home and abroad.Explore the development trend of next point-of-interest recommendation,and provide solid theoretical support for the follow-up research work.(2)Analyze the time and distance characteristics of the user's historical check-in se-quence.In this paper,we propose a new Spatio-Temporal Gated Network by enhancing Long Short-Term Memory network,where spatio-temporal gates are introduced to capture the spatio-temporal relationships between successive check-ins.Specifically,two pairs of time gates and distance gates are designed to control the short-term interest and the long-term interest updates,respectively.Moreover,we introduce coupled input and forget gates to reduce the number of parameters and further improve efficiency.Finally,we evaluate the proposed model using four real-world datasets from various location-based social networks.The experimental results show that our model significantly outperforms the state-of-the-art approaches for next point-of-interest recommendation.(3)Under the framework of deep learning,using semi-supervised learning,supplement-ing with context information to alleviate data sparsity in the case of insufficient user check-in data.A recommendation method for joint semi-supervised context map and spatio-temporal gated network is proposed to joint learning the point-of-interest embedding vector.The geographic context and the user's preference of the point-of-interest are merged to further improve the performance of the next point-of-interest recommendation.Then the effective-ness and reliability of the proposed method are verified by experiments.