Research On Point Of Interest Recommendation Based On Neural Network

In recent years,the development of the mobile Internet has changed people’s live and work,making them efficient and convenient.In terms of tourism travel,people can obtain relevant travel service information through the Internet.However,the growth of online users will lead to a large amount of data generated on the Internet.It is extremely inconvenient for users to spend a lot of time inquiring related travel information.To address the problem of information overload,the recommendation system can effectively filter redundant information,and provide users with satisfactory recommendation services for POI(Points of Interest).In the field of travel recommendation,traditional POI recommendation methods have achieved effective recommendation list,but these methods that rely on the design of shallow features cannot comprehensively learn the deep features of users and POIs.Traditional recommendation models also suffer sparse data and the problems of recommendation inefficiency,which can seriously hinder the performance of POI recommendation result.This paper uses POI recommendation based neural network to deal with the above challenges.This paper provides POI recommendation for users by analyzing users’ preferences based on knowledge graph and context information and improves the efficiency of the recommendation.The main content of paper is as follows:1.Aiming at the problem that the in-depth features of users and attractions cannot be effectively learned and the problem of features of the attractions’ features cannot be properly trained in the training process in the personalized recommendation methods,which can result in unsatisfactory recommendation list.In this paper,we propose a model that fuses knowledge graph and neural network for tourist attraction recommendation.In our proposed model,the embedded representation of the attraction feature sub-graph is accurately obtained by the knowledge graph embedding.Then the attraction vector generated by fusing the features of attraction is combined with user vector to further mine the deep features of the user-attraction interaction in the neural network,and get more accurate recommendation.Through theoretical analysis and experimental results,our proposed model is demonstrated to be effective.2.Aiming at the issues of existing POI model suffer sparse data,cannot well distinguished the influence of user’s friends on user preferences,and ignoring the decision of user caused by the features of POI.We present a model that exploiting context influence for POI recommendation.Our presented model first obtains the vector representation of userthrough the user’s context module,which calculates the different weights of the interests and overall familiarity between user and user’s friends.Secondly,the influence of geographic distance and POI’s latent feature is calculated by the POI’s context module that can obtain the POI vector relating to the user’s preferences.Finally,the user vector and the POI vector are combined to predict the POIs that user has been unvisited.Compared with the benchmark methods,the performance of our presented model is demonstrated to significantly improve.