The Application Of Multi-Criteria Recommendation Algorithm Considering Social Relationships And Group Preferences In Tourism E-Commerce Platforms

The rapid development of the tourism industry has enriched people's daily life,but the excessive and inferior information on the tourism e-commerce platform have caused problem to people's decision-making.In this context,recommender system,as an information filtering technology,plays an increasingly important role in e-commerce platforms.The precise and efficient recommender systems deployed on the traditional tourism e-commerce platforms can optimize the user experience,improve the user purchase conversion rate,and promote the benign development of the platforms.Recently,the criteria ratings as supplemental information has been shown to significantly improve recommendation quality.However,due to the expansion of rating dimension,most multi-criteria recommendation algorithms have poor scalability and are not suitable for large online travel e-commerce websites.Moreover,the current multi-criteria recommender systems still has potential for improvement in accuracy.Therefore,the main research contents of this paper include the following points:Firstly,under the trend of social tourism e-commerce platforms,this paper analyzes the classification,features and differences of user social relationships on platforms,and discusses how to use multi-criteria ratings to construct the implicit social relationships when explicit social relationships is missing.Secondly,in order to get more accurate recommendation results and alleviate the cold start problem,this paper proposes an extended social recommendation model TPSVD.This model considers the transitivity of the user's tastes and the implicit influence of social information and ratings information.Next,this paper proposes a new multi-criteria recommendation framework,which first uses TPSVD to predict the criteria ratings,and then uses SVR to predict the overall ratings.In order to improve the accuracy and scalability of the multi-criteria recommendation algorithm,the clustering technology is used to find the user groups and groups' preference information.Finally,through the data sets from TripAdvisor,the proposed models and other models in this paper are compared.The results show that the proposed models have advantages in predicting performance and recommending performance.