Studies On The Forecast Of Unsafe Events For Airlines

Through studying on the forecast of the civil aviation unsafe events, finds that most of the forecast is based on the history data of the unsafe events, they set up math models to forecast the future data. The unsafe events of airlines are influenced by many factors, and the influencing factors are changing. First analyzes the influencing factors, and discusses several main factors with principal component analysis algorithm, rejecting the factors with minimum weight. The five main factors are resource management, crew resource management, skill-based error, inadequate supervision and planned inappropriate. Put the five main factors as the input stylebook of the airlines'unsafe events forecast model. The airlines'unsafe events contain accidents, incidents, severe errors and common errors. Put the four types of the unsafe events as the output stylebook of the forecast model.In order to solve the BP neural network's inherent deficiency of slowly converging and easily falling into local minimum, we propose to use an algorithm combined self-adapting learning rate and extra momentum. According to the data of an airline , we adopt a three-layer BP neural network. Input-layer contains five nerve cells and output-layer contains four nerve cells. Put the data form January to September as the train stylebook, the data of October and November as the test stylebook. The result shows the feasibility of the forecast model and it is valuable. Input the data of input-layer and output-layer from January to November, use every parameter of the model to forecast, when input the main influencing factors of December, we can forecast the unsafe events of December. The result shows that the unsafe events'number of December is more than that of November. It gives the workers a caution. The workers find the fluencing factors and give advices to prevent the unsafe events.