基于简化WITI指标的机场延误预测方法

Airport Delay Prediction Method Based on Simplified Weather Impacted Traffic Index

为实现在战略或预战术阶段对恶劣天气条件下的机场延误进行有效预测,本文引入简化的天气影响交通指标(WITI),采用灰色关联分析的方法,验证该指标与机场延误之间的关联性,再分别以WITI指标和传统指标构建多元线性回归模型和BP神经网络预测模型,对广州白云国际机场(ZGGG)和深圳宝安国际机场(ZGSZ)的离场延误进行预测.结果显示,WITI指标与机场离场延误之间的关联度明显高于传统指标,基于WITI指标比基于传统指标构建的多元线性回归模型,在预测准确度上高出14.09%(ZGGG)和9.79%(ZGSZ),同样在BP神经网络模型中则高出8.00%(ZGGG)和6.41%(ZGSZ),由此认为WITI指标在机场延误预测中具有更好的应用效果.

This paper aims at predicting airport delay accurately in the strategic or pre-tactical stage. We introduce the simplified Weather Impacted Traffic Index（WITI）, and use grey incidence analysis method to verify the correlation between the WITI and actual airport delay. Then we use WITIs and traditional indexes to develop multiple linear regression model and BP neural network predictive model respectively, and we chose the data of Guangzhou Baiyun International Airport（ZGGG） and Shenzhen Baoan International Airport（ZGSZ） as research samples. According to the result, WITIs are more closely related to airport departure delay than traditional indexes. Furthermore, the predicted accuracy of the multiple linear regression model based on WITIs increases 14.09%（ZGGG） and 9.79%（ZGSZ）, compared with the model based on traditional indexes. When the multiple linear regression model was replaced by the BP neural network predictive model, the results increase 8.00%（ZGGG） and 6.41%（ZGSZ） as well. Thus the WITI has a comparatively satisfactory feedback in airport delay forecast.