基于LightGBM和SHAP的可解释性航班延误预测

打开文本图片集

中图分类号：V351 文献标识码：A 文章编号：1006-8228（2025）11-26-06

Interpretable Flight Delay Prediction based on LightGBM and SHAP

Cheng Xiaokang，Huang Ruoqian

（CivilAviation Flight University of China，Guanghan，Sichuan 6183o7，China）

Abstract：Toadressthechallengesofhyperparameteroptimizationandquantitativeanalysisoffeaturesynergyinflightdelay prediction，thispaperproposesaninterpretablepredictionmethdtatintegrateseficientparametersearchandinteractioeffect analysis.ByconstructingthemodelwithLightGBManddesigninganadaptiveoptimizationstrategy，thehyperparametersearch eficiencyisimprovedFurthemoretheSHAPInteractionItesityCoeficient（SHAP-I）isintroducedtoquantifytheglobal synergystrengthbetweenfeaturepairs，overcomingthelimiationoftraditionalmethodsthatonlyexplainthemainefectsof features.Experimentson2022datafromanairportdemonstratethatthemodelsignificantlyimprovespredictionacuracyand SHAP-ISefectivelyidentifeskeysynergythresholdsamong"traffcvolume，historicaldelays，andresources"，providinga quantitative decision-making basis for dynamic airport control.

Keywords：Flight Delay；Lightweight Gradient Boosting；ExplainableAI；Feature Interaction

0引言

航班延误是影响民航运行效率和乘客体验的核心问题。（剩余7021字）