Identification of Risk Issues in Civil Aviation Operation Units

According to the own characteristics of civil aviation operation unit data, it is important to apply deep learning technology to identify the data of civil aviation operation risk problems. In this paper, we first collect data from relevant operation departments and manually annotate the risk issues in the data to form a text library for risk issue identification. And data pre-processing is carried out to analyze the data visually byword frequency and co-occurrence word analysis. Second, text splitting is performed on the basis of expanding the basic corpus. Considering the characteristics of words such as polysemy, the corresponding algorithm is studied and the words are vectorized using Word2vec method. Then, a neural network model incorporating LSTM and Bi-LSTM models was built for risky problem identification. Techniques such as dynamic learning rate correction and writing callback functions are used to learn the corresponding model parameters with the aim of reducing generalization errors. Finally, the above-mentioned models and algorithms are integrated, and the model optimization and reconfiguration finally form the risk problem identification technology for civil aviation operation. The practical application results are good, and the technique can be extended to the identification of civil aviation operation risk problems.