Research On Classification Method Of Civil Aviation Uncivilized Passengers Based On Deep Learning

In recent years,the uncivilized behavior of civil aviation passengers has threatened the safety of air transportation in our country.The classification of hazard levels for these uncivilized passengers is an important measure to maintain flight safety.Through dynamic risk assessment of passenger behavior records,different security inspection measures are adopted for passengers with different risk levels,which not only ensures strict inspection of high-risk passengers,but also reduces the time for low-risk passengers to queue for security inspection.The behavior descriptions and public opinion records of uncivilized passengers are all text data.In order to complete the identification of potential uncivilized passengers,the text is first classified into two categories..In order to solve the problem of low accuracy of text classification.this paper first proposes a combination model of sparse self-attention mechanism and Bi LSTM.In this model,Cw2 vec was used to obtain Chinese word vectors,and then Bi LSTM and Sparse Self Attention were used for feature extraction.The experiment was conducted on the public hotel review data,and the accuracy reached 91.96%,which provided a method basis for the identification task of potential uncivilized passengers.Secondly,through multiple text classification,the identified civil aviation uncivilized passengers are classified into risk levels,and a classification method of civil aviation uncivilized passengers based on confrontation training and Bi LSTM-CNN is proposed.Firstly,adversarial training is added to the word embedding layer to improve the robustness of the model.Then,Bi LSTM is used to extract the key contextual information of the network public opinion.Finally,multi-channel convolutional neural network is used to extract the local text features again.The experimental results show that the accuracy of the method based on confrontation training and Bi LSTM-CNN reaches 80.52%,which is better than other deep learning models.