Research On Civil Aviation Passenger Overlapping Community Discovery Based On Complex Network

With the explosive growth of civil aviation passengers,airline passenger data is characterized by massiveness and complexity.Complex Network are widely applied because of their characteristics of exposing the essence of complex systems,discovering hidden laws,and predicting network evolution.This paper aims to mine civil aviation passengers overlapping community structure based on Complex Network,which has important practical application for airlines to develop revenue-enhancing marketing strategies.In this paper,we propose a civil aviation passenger relationship recognition model(CAPR-RM)based on generating adversarial networks,which is mainly concerned on the problem of low recall rate of passenger relationship recognition in the construction of civil aviation passenger relationship network.Firstly,data fusion is performed to extract common travel characteristics between passengers.Secondly,we design a Generative Adversarial Network(GAN)to adopt the real passenger relationship.Finally,the passenger relationship is identified by the discriminant function to calculate the distance between the processed passenger relationship data and the original data.The experimental results on the real passenger service dataset show that our CAPR-RM model improves the recognition accuracy,recall rate and F1 value to above 90%.A multi-tag propagation overlapping community discovery method(PCMLPA)that introduces pairwise constraints is proposed,which mainly focuses on the problem of poor robustness and low accuracy of traditional label propagation algorithms.Firstly,it integrates explicit with implicit double-layer passenger relationships to build a civil aviation passenger relationship network.Secondly,by combining with the characteristics of low cost and constraint expansion of paired constraints in civil aviation passenger relationship network,the semi-supervised learning is introduced and the constraint expansion rules in overlapping communities are designed to improve the accuracy of community detection and utilization efficiency of constraints.Finally,the order of node update and traversal during tag propagation is optimized to enhance the robustness of the algorithm.Experimental results on synthetic networks and real network datasets show that the PCMLPA method has better robustness and time complexity.In terms of the accuracy of community detection results,the method introduces a phase with a constraint of 5%.It is 13.46% higher than other algorithms and is significantly better than other overlapping community discovery algorithms.