基于SMOTE算法的深度神经网络航班延误预测模型  

Flight Delay Prediction Model of Deep Neural Network Based on SMOTE Algorithm

作　　者：郑玉帆 徐海文[1] ZHENG Yu-fan;XU Hai-wen(Civil Aviation Flight University of China,Guanghan 618307,China)

机构地区：[1]中国民用航空飞行学院,广汉618307

出　　处：《价值工程》2023年第4期118-120,共3页Value Engineering

摘　　要：航班延误预测在民航运输降低延误成本、规划航班计划方面有重要作用。针对航班延误数据不平衡的问题,本文提出一种基于SMOTE算法的深度神经网络航班延误预测模型(SMOTE-DNN)。该模型首先利用过采样技术SMOTE算法对原始数据进行处理,减小不平衡数据对模型的影响;接着利用深度神经网络实现航班延误等级预测。此外,本文将所提模型用于真实数据的航班延误预测。结果表明,对航班延误失衡数据进行平衡处理后可以提高模型的拟合度和预测精度,SMOTE-DNN的预测精度可以达到88.79%,且对航班延误各等级均有着较好的预测效果。Flight delay prediction plays an important role in reducing delay costs and planning flight schedules in civil aviation transportation. For the imbalanced data of flight delay, a flight delay prediction model based on the SMOTE algorithm and deep neural network(SMOTE-DNN) is proposed. First, SMOTE-DNN uses the oversampling technique, SMOTE, to process the original data to reduce the influence of imbalanced data. Second, the deep neural network is used to predict flight delay classification. In addition, the proposed model is used to forecast with real data. The results show that the fitting and prediction accuracy of the model can be improved after balancing the flight delay imbalanced data. The prediction accuracy of the SMOTE-DNN can reach 88.79%, and it has a good prediction effect on all classifications of flight delay.

关 键 词：SMOTE算法 深度神经网络 航班延误 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]