Research On Information Matching Of Uncivilized Passengers In Civil Aviation Based On Fusion Similarity

The rapid development of the air transport industry has gradually increased the number of people traveling on civil airliners.However,the violations of uncivilized passengers in civil aviation affected aviation safety during the flight.Civil aviation industries in various countries have established uncivilized passenger databases to effectively identify uncivilized passengers and develop corresponding service measures for them.Therefore,in order to predict the potential uncivilized level of civil aviation passengers and effectively identify and monitor uncivilized passengers,it is urgent to establish a quantifiable and measurable index system to analyze the similarity of uncivilized behavior of civil aviation passengers,and to realize the effective identification of uncivilized passengers.Firstly,a key information extraction method based on Dual-encoder has been proposed to build a regularized,standardized data set,and to build a behavior rule base of uncivilized passengers,which integrates non-homologous sources from multiple platforms such as civil aviation,railway,tourism,and the internet heterogeneous data.This method combines the sequence mapping model Seq2 Seq with attention and adds TF-IDF,part of speech(POS),and key score(Sco)features to enhance word vector representation.At the same time,according to the matrix perturbation theory,the optimal dimension of word embedding is optimized to make full use of the ability of word vector representation.The experimental results show that the proposed method in the Rouge evaluation system is as high as 0.39,which is 0.1-0.13 higher than other methods.It can be seen that the method proposed in this paper has higher extraction accuracy and effectively solves the problem of difficult extraction of key information in the face of complex texts.Secondly,a model of uncivilized passenger’s behavior similarity matching based on joint similarity is proposed to enhance the matching accuracy of the similarity of uncivilized passengers and improve the prediction accuracy of the potential degree of uncivilized passengers.This method combines word vector decision weighting algorithm to generate sentence vectors for behavior records.And,the punishment rules and punishment results of passengers are incorporated into the calculation model.The multi-granularity measurestandard of passengers is designed based on the social distribution law of uncivilized passengers.The multi-dimensional characteristics of passenger behavior and punishment results are integrated for joint similarity calculations,and to predict the potential uncivilized passengers in engaging in civil aviation activities.The experimental results show that the accuracy of the proposed method to predict the potential uncivilized level of civil aviation uncivilized passengers is as high as 82.25%,which effectively improves the ability of passenger identification and the accuracy of level prediction.