The Study on the Prediction Models of Human Factor Flight Accidents by Combining Fuzzy Clustering Methods and Neural Networks

摘要

This study was meant to effectively prevent and reduce the occurrencernof fatal accidents. The method used herein combined the multi-fuzzyrnC-means clustering algorithm and the BPN-based neural network were usedrnto improve both correlation and accuracy. The hybrid prediction of neuralrnnetworks was implemented using two groups of prediction-training models.rnThis hybrid method had a good prediction effect on flight accidents since itsrnaccuracy is greater than that of general statistical models. The resultsrnshowed that the Mean Absolute Error (MAE) and Mean AbsoluternPercentage Error (MAPE) were rated at 10.2639 and 12.5258%,rnrespectively, and the first group exerted the best performance whenrnaccuracy was combined with FCM-BPN prediction. For the FCM-BPNrnmodels vested with the grouping concept, the research results indicated thatrnthe prediction accuracy was significant and could appropriately andrneffectively monitor the risk management of flight safety made by airlinesrnand flight management agencies.