| With the rapid development of the Internet and the increasingly strong personalized demands for air services,the recommendation methods of air passenger services have attracted more and more scholars’ attention.However,with the continuous enrichment of civil aviation passenger services,the problem of information overload is becoming more and more serious.At present,the main problems faced by civil aviation passenger service recommendation are the accuracy of personalized active recommendation of civil aviation passenger service and the sparseness of data.The interactive data of passenger and aviation project constitutes the structure of the diagram,and the graph neural network technology has good nonlinear expression ability and implicit feature mining ability.Therefore,we solve the related problems in the civil aviation passenger service recommendation system by drawing neural network technology and combining the theory of social network relationship,which is of great significance to the civil aviation passenger service recommendation.In view of the problem that the accuracy of personalized active recommendation of civil aviation passenger service items is not high,we first put forward a model of implicit interactive feature mining of civil aviation passengers based on graph neural network.First of all,on the basis of the GNN of the recommended field of civil aviation passenger service project,the finegrained modeling and design scoring rules of civil aviation passenger service projects are introduced.Then,via using the graph neural network to excavate the implicit interaction characteristics between passengers and fine-grained properties and embed them into the entity,the potential embeddings of the valid passenger and fine-grained f civil aviation passenger service item properties are obtained.Finally,the civil aviation passenger service project is recommended by the comprehensive fine-grained attribute score,which produces a more accurate and explanatory list of recommendations.Through empirical research on the actual airline operation data,the experimental results show that the evaluation index RMSE of the implicit interactive feature mining model of civil aviation passengers based on the graph neural network is improved by about 3.5% compared with the traditional model on average,and the significant advantages of the model in score prediction are verified.In addition,in order to solve the problem that the interactive data between passengers and civil aviation service items is too sparse,we put forward a recommended model for mining the characteristics of civil aviation passenger relations based on the social relationship network.First,the social relationship network is constructed by historical order data,and then the graph neural network is constructed based on the network,which is used to excavate the implicit preference characteristics of the passenger’s social relations and integrate with the passenger entity to obtain the potential character of the final passenger’s implicit characteristics,and improve the accuracy of the model recommendation.Through comparative experimental analysis,the effectiveness of the introduction of social relations network is verified,and the evaluation index is improved by about 4.53% compared with the traditional algorithm,which has improved the accuracy of the recommendation of civil aviation passenger service items. |
