Research On Civil Aviation Passenger Behavior Modeling For Airline-Route Selection

Analyzing the civil aviation passenger travel airline-route selection behavior and discovering its inherent mode to predict possible future airline-route selection behaviors,it not only helps airlines to implement personalized travel product recommendations for different passengers to improve passenger satisfaction,but also helps airlines making route planning decisions through finding group travel patterns based on individual travel needs.This paper focuses on the modeling of civil aviation passenger behavior of airline-route selection.The specific work is as follows:Based on the traditional recommendation algorithm based on user collaborative filtering,an airline-route selection behavior prediction model based on passenger similarity is proposed.The two methods are used to find the target by calculating the similarity between passenger vector and the correlation between passenger correlation networks.The passenger's similar passenger set,combined with the passenger's own preferences and similar passenger preferences,is used to predict the passenger's possible future airline-route selection behavior.The experimental results on the two-year booking log data set of civil aviation passengers show that the airline-route selection behavior prediction model based on passenger similarity has higher precision and recalling than the traditional forecasting methods based on airline-route heat and history selection.Considering that the passenger's future airline-route selection behavior is not only affected by its own preferences and similar passenger preferences,but also by the correlation between routes,an airline-route selection behavior prediction model based on multiple correlations is proposed.Based on the civil aviation passenger booking log constructs a passenger-to-airlineroute selection bipartite graph network,we construct a passenger-related relationship network and a route-related relationship network by calculating passenger correlation and route correlation,and adopt node2 vec in the related relationship network.The nodes are vector represented to solve the high dimensional sparsity problem.The experimental results on the two-year booking log data set of civil aviation passengers show that the airline-route selection behavior prediction model with multiple correlations has higher precision and recalling than the airline-route selection behavior prediction model based on passenger similarity.