基于深度神经网络的航班保障时间预测研究

航班地面保障时间预测是提高机场运行保障效率和决策能力的关键问题之一。考虑到服务流程的复杂性和特殊性,建立了航班地面保障资源到位时间的高斯概率模型,提出了一种基于深度神经网络的航班地面保障时间预测模型,并根据保障数据规律性变化调节模型参数,减小不确定性因素产生的泛化误差。研究结果表明,单航班预测结果的平均绝对误差比多航班小4.479 min,模型评价分数达到了94.608,且预测精度比传统BP神经网络和贝叶斯网络方法高3%~5%。

Research on Flight Ground Service Time Prediction Based on Deep Neural Network

Flight ground service time prediction is one of the key issues in improving the airport operational efficiency and decision making capacity. Taking into account the complexity, particularity and uncertainty of the service process, a Gaussian probability model of flight ground service resource in place time is established, a flight ground service time prediction model based on the deep neural network is proposed. According to the regular changes of operational data, the model parameters are adjusted to reducet the generalization error caused by other factors. The research results show that the average absolute error of time prediction under single flight is 4.479 min less than that of the multiple flights, the model evaluation score reaches 94.608, and the prediction accuracy is 3%~5% higher than that of the traditional BP neural network and Bayesian network method.