摘要
为了提高对机场航班延误时间的准确性,对预测模型进行了研究。采用麻雀搜索算法(SSA),结合卷积神经网络(CNN)和双向长短时记忆网络(BiLSTM),提出了一种基于SSA-CNN-BiLSTM的航班延误预测模型。使用美国亚特兰大机场的实际运行数据进行了验证,与BiLSTM,CNN-LSTM等基准模型进行了比较试验,并加入流量和时间双特征数据集验证模型性能。结果表明,SSA-CNN-BiLSTM模型在评价指标上表现最优,其平均绝对误差(MAE)为5.15,均方根误差(RMSE)为7.58,预测精度优于基准模型,具有良好的多特征处理能力。
In order to improve the accuracy of flight delay time at airports,the prediction model was investigated and a flight delay prediction model based on SSA-CNN-BiLSTM was proposed by using Sparrow Search Algorithm(SSA)in combination with Convolutional Neural Network(CNN)and Bidirectional Long and Short Term Memory Network(BiLSTM).Validation is carried out using actual operational data from Atlanta Airport in the U.S.A.Comparison tests are conducted with benchmark models such as BiLSTM,CNN-LSTM,and the model performance is verified by adding the dual feature dataset of flow and time.The results show that the SSA-CNN-BiLSTM model performs optimally in the evaluation indexes,with a mean absolute error(MAE)of 5.15 and a root mean square error(RMSE)of 7.58,and the prediction accuracy is better than that of the benchmark model,with good multi feature processing capability.
作者
杨新湦
游超
YANG Xin sheng;YOU Chao(Civil Aviation University of China,Tianjin 300000,China)
出处
《航空计算技术》
2024年第5期22-26,共5页
Aeronautical Computing Technique
基金
国家自然科学基金与民航基金联合重点项目资助(U2133207)。
关键词
航班延误预测
参数优化
卷积神经网络
双向长短时记忆网络
麻雀搜索算法
flight delay prediction
parameter optimization
convolutional neural networks
bidirectional long and short term memory networks
sparrow search algorithm
