Published October 3, 2026 | Version v1
Publication Open

A Residual Embedding Network for Flight Departure-Delay Forecasting — Deep Learning Edition

Authors/Creators

Description

SkyPath Delay Desk estimates the probability that a scheduled US flight departure will leave more than 15 minutes late. This deep learning edition rebuilds the original LightGBM project around a PyTorch residual embedding network, and adds calibrated probabilities, a model-uncertainty band on every forecast, and a fuller evaluation section.

The task, inputs and target are unchanged from the original project: 25 pre-departure inputs across seven groups (calendar, flight, airport traffic, airline, aircraft, location and weather) feed a binary classifier for DEP_DEL15, whether a flight left 15 minutes or more late. Columns known only after departure are excluded to avoid leakage.

Live app: https://skypathdl-prediction.streamlit.app/

Files

Files (28.9 kB)

Name Size
md5:4ea0e814110fdfb05bd1a8663f3d0b3e
28.9 kB Download

Additional details

Software

Repository URL
https://github.com/ShlokP2006/skypath_dl.git
Programming language
Python
Development Status
Active

References

  • threnjen. 2019 Airline Delays w/Weather and Airport Detail [data set]. Kaggle. https://www.kaggle.com/datasets/threnjen/2019-airline-delays-and-cancellations
  • Gorishniy, Y., Rubachev, I., Khrulkov, V., and Babenko, A. (2021). Revisiting deep learning models for tabular data. Advances in Neural Information Processing Systems 34.
  • Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735 1780. https://doi.org/10.1162/neco.1997.9.8.1735
  • Guo, C., Pleiss, G., Sun, Y., and Weinberger, K. Q. (2017). On calibration of modern neural networks. Proceedings of the 34th International Conference on Machine Learning.
  • Gal, Y., and Ghahramani, Z. (2016). Dropout as a Bayesian approximation: representing model uncertainty in deep learning. Proceedings of the 33rd International Conference on Machine Learning.
  • He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recognition. IEEE CVPR.
  • Loshchilov, I., and Hutter, F. (2019). Decoupled weight decay regularization. International Conference on Learning Representations.
  • Ke, G., et al. (2017). LightGBM: a highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems 30.