基于分类回归决策树算法的航班延误预测模型  

Flight delay prediction model based on CART algorithm

作　　者：王辉[1] 张文杰 刘杰[1] 陈林烽 李泽南 WANG Hui;ZHANG Wenjie;LIU jie;CHEN Linfeng;LI Zenan(College of Aeronautical,CAUC,Tianjin 300300,China;Tianjin Artificial Intelligence Innovation Center,Tianjin 300300,China)

机构地区：[1]中国民航大学航空工程学院,天津300300 [2]天津(滨海)人工智能军民融合创新中心研究一部,天津300300

出　　处：《中国民航大学学报》2022年第3期35-40,共6页Journal of Civil Aviation University of China

基　　金：国家自然科学基金项目(U1733128)。

摘　　要：针对民航客机航班延误问题,构建了基于随机森林(random forest)与分类回归决策树(CART,classification and regression tree)算法的航班延误预测模型,利用国内大型机场的真实数据集对模型进行训练,通过与Logistic回归算法,K-近邻回归(KNN,K-nearest neighbor)算法和决策树(decision tree)算法的训练结果对比,从拟合效果可以看出,该方法可以处理高维度数据,泛化能力好,降低了过拟合的可能性,模型的拟合程度R2可以达到0.83。Aiming at the flight delay problem of civil aviation airliners,this paper constructs a flight delay prediction model based on random forest model and classification regression decision tree(CART)algorithm,on which a large amount of training are conducted by using the real data set from large China airports.With the training results of Logistic regression algorithm,K-nearest neighbor(KNN)regression algorithm and decision tree algorithm,and with the fitting results,it is concluded that this method can deal with high latitude data with good generalization ability,and reduce the possibility of over fitting.The fitting degree R^(2) of the model can reach 0.83.

关 键 词：航班延误 随机森林模型 分类回归决策树(CART)算法 

分 类 号：V355.1[航空宇航科学与技术—人机与环境工程]