Predicting Flight Delays at Soekarno-Hatta Airport Using Machine Learning: a Comparative Analysis of Random Forest and Gradient Boosting with SMOTE

This research aimed to perform a comparative analysis of Random Forest and Gradient Boosting models with SMOTE. The study evaluated these two models to assess their performance. Applied upsampling techniques, such as SMOTE (Synthetic Minority Over-sampling Technique), to enhance their performance, with particular emphasis on investigating how upsampling improved predictions, especially for unbalanced data. The models were thoroughly trained and tested, both with and without upsampling, for a comprehensive comparison. The results showed that both Random Forest and Gradient Boosting models exhibited improved accuracy with upsampling, indicating their effectiveness in dealing with flight delays at Soekarno-Hatta Airport based on a dataset of 1.5 million records flight from 2018 to 2022. From the research, the Gradient Boosting model with SMOTE stood out with the highest values, showcasing its performance: Accuracy (Acc): 91.17%, Area Under the Curve (AUC): 0.97, Precision (Prc): 91.15%, Recall (Rec): 91.20%, F1-S core (F1-S): 91.18%, and Processing Time: 01:43 minutes. These results had implications for improved planning and passenger satisfaction, demonstrating progress in the prediction of flight delays.