Research On Personalized Tourism Recommendation Algorithm Based On Social Network

Tourism is an important industry.In recent years,a large number of tourists have traveled in the form of self-help tourism.Automatically recommending points-of-interest(POI)and tourist routes that meet the individual needs of tourists is the key to personalized tourism recommendations.The widespread use of mobile smart devices has formed a Location-based Social Network(LBSN).The increasingly number of users share their "check-in" records or travel photos in social networks,generating a large amount of social network data containing abundant user travel time and space information and contextual information,which provides excellent opportunities to understand user behavior and implement personalized travel recommendations.In this paper we focuses on the key issues of personalized travel recommendation: POI recommendation and personalized travel route recommendation.In order to mine users' personalized preferences effectively during the travel process,we propose a POI recommendation algorithm and a travel route recommendation algorithm based on deep learning.The main contents of this paper are as follows:(1)Feature analysis and vectorization of social network data.We determined topic features,geographic factor features,and user behavior characteristics' effectiveness by features analysis.In the vectorization of experimental data,we define a novel topic model to perform vectorization processing;we use normalization methods to vectorize geographic factor features;and we vectorize user behavior features according to Matrix Factorization(MF)algorithms.The vectorization of the features alleviate the sparseness of the social network data.(2)We proposed a POI personalized recommendation algorithm based on deep neural network.This model utilizes feature embedding technology to integrate the effective features in social networks,uses deep learning models to extracts high-level interactions between features.According to the verification on the real experimental data set,it obviously shows that the proposed algorithm has significantly higher metrics than other comparison algorithms.Among them,the accuracy of the recommended POIs in number of the top five,the top ten,and the top twenty increase 9.9%,7.4% and 7%,the highest recall rates increase 4.2%,7.5% and 14.4%.(3)We propose a personalized travel route recommendation algorithm based on users' dynamic interests.By weighting the user's historical travel interest and target area characteristics,we mine the user's dynamic interests during the travel process to recommend travel routes to the user.Through experimental verification on a real experimental data set,the proposed method has significantly higher metrics in the recommendation algorithm than other contrasted personalized travel route recommendation methods,and is closer to tourists' real travel routes and more practical.To sum up,this paper proposes a personalized POI recommendation,which effectively integrates multiple features in social networks,and utilize deep learning technology to effectively alleviate data sparseness and feature learning problems in travel recommendation.The research has important usage scenarios and business value in the tourism industry.