基于弹性神经网络的航班延误时间预测

Flight Delay Prediction Based on Elastic Neural Network

航班延误受到多种因素的交叉影响,导致航班延误数据分布不规律,难以从传统统计学的角度准确预测航班延误时间,因此以减少数据过拟合为目标,利用随机森林特征选择模型筛选21个重要特征,引入正则化L1、L2范数,建立弹性神经网络预测模型,对航班落地延误时间进行预测。预测结果为:±3 min容差内的准确率达到83.954%,±5 min容差内的准确率达到94.431%,结果表明该模型能够提高航班延误预测的准确率。

Flight delays are affected by multiple factors,resulting in irregular distribution of flight delay data,and it is difficult to accurately predict flight delay time from the perspective of traditional statistics.Therefore,with the goal of reducing data over fitting,21 randomized forest feature selection models are used to screen.The important feature is to introduce the regularized L1 and L2 norms,establish an elastic neural network prediction model,and predict the flight delay time.Finally,the accuracy within ±3 minutes tolerance is 83.954%,and the accuracy within ±5 minutes tolerance is 94.431%.The results show that the model can improve the accuracy of flight delay prediction.