Neural Network Based Aircraft Automation Early Warning Research

Abstract

With the accelerated urbanization of China and the rapid growth of urban economy, the scale of China's civil aviation industry continues to expand and the civil aviation industry continues to flourish. However, the resulting safety problems and exposed safety hazards are also gradually increasing. In this paper, an automated aircraft warning model based on random forest and bp neural network is studied, and an effective scheme is given based on different flight crews, flight routes, airports, and flight records under specific flight conditions. Firstly, data pre-processing is performed to get 156 features. The important features are filtered using random forest model and the top 10 most important features are taken. A neural network model was built and after normalizing the data, the remaining 10 most relevant flight parameters were used to evaluate the pilot's flight technique, and 100% training accuracy and 79% validation accuracy were obtained. The results were then validated using support vector machines, decision trees, and plain Bayesian models, respectively, with the final neural network predicting the best. In addition, we also designed an automated aircraft warning model with a combination of random forest and neural network. The loss of which is about 0.095, and the test performance is better than the general neural network model and random forest model in the test set.