Estimate weather impacted airport capacity using ensemble learning

Ensemble BDT consistently outperforms the single SVM classifier. This demonstrates that multiple classifier systems are more robust in the presence of noise and other imperfections in data as compared to a single classifier system. • The BDT classifier provides very good estimates of the runway configuration using the airport weather. • The AAR classification predictions by BDT for 2 and 4 hour look-ahead times are excellent. For 6-hour AAR prediction, the performance of the BDT classifier is not bad, AUC is 86% for EWR and 92% for ORD. • The AAR prediction results using BDT models for EWR are not as good as for ORD (Weather factors).