arxiv: 2604.20288 · v1 · submitted 2026-04-22 · 💻 cs.LG

Recognition: unknown

Generative Augmentation of Imbalanced Flight Records for Flight Diversion Prediction: A Multi-objective Optimisation Framework

Karim Aly , Alexei Sharpanskykh , Jacco Hoekstra

Authors on Pith no claims yet

Pith reviewed 2026-05-10 00:42 UTC · model grok-4.3

classification 💻 cs.LG

keywords flight diversion predictiongenerative data augmentationimbalanced datasetsmulti-objective optimizationdeep generative modelsaviation machine learningsynthetic tabular datapredictive modeling

0 comments

The pith

Optimised generative models create synthetic diversion records that substantially improve prediction accuracy over real flight data alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper addresses the scarcity of flight diversion records by exploring how generative models can produce synthetic examples to balance training data for machine learning predictors. It introduces a multi-objective optimisation framework paired with hyperparameter search to tune three deep generative models—TVAE, CTGAN, and CopulaGAN—alongside a Gaussian Copula baseline. The synthetic data undergoes a six-stage evaluation covering realism, diversity, statistical similarity, and predictive utility. Results indicate that augmentation with the optimised synthetic records yields higher diversion prediction performance than training on the original imbalanced records. This approach offers a practical route to better models for rare high-impact events in aviation without needing vastly more real observations.

Core claim

A multi-objective optimisation framework with automated hyperparameter search, applied to deep generative models such as TVAE, CTGAN, and CopulaGAN, generates synthetic diversion records whose quality enables diversion prediction models to achieve substantially higher accuracy than models trained solely on historical real flight records.

What carries the argument

Multi-objective optimisation framework with automated hyperparameter search that configures generative models to produce high-quality synthetic tabular aviation data.

If this is right

Optimised generative models outperform their non-optimised versions across realism, diversity, and utility metrics.
Augmenting real records with synthetic diversions raises predictive accuracy for rare diversion events.
The six-stage evaluation framework provides a structured method to validate synthetic data for operational aviation use.
Deep generative approaches can surpass statistical baselines such as the Gaussian Copula for this task.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The framework could be tested on other rare-event tabular datasets in transportation or healthcare to check broader applicability.
Real-time integration of such augmented models might support proactive diversion risk alerts in flight operations.
Cross-airline validation would clarify whether the optimisation choices transfer to different operational environments.

Load-bearing premise

The synthetic diversion records produced by the optimised generative models are sufficiently realistic and unbiased that they improve real-world predictive performance rather than introducing artifacts.

What would settle it

A held-out test set of real flight records where a model trained on the augmented data shows no improvement or lower accuracy than a model trained only on real data would falsify the central claim.

Figures

Figures reproduced from arXiv: 2604.20288 by Alexei Sharpanskykh, Jacco Hoekstra, Karim Aly.

**Figure 1.** Figure 1: Overview of the analysis framework. 3.1. Data and Preprocessing This study relied on the publicly available “TranStats Database for Airline On-Time Performance” provided by the Bureau of Transportation Statistics (BTS) (Bureau of Transportation Statistics, 2023; U.S. Department of Transportation). The database covers U.S. domestic flights and contains comprehensive records on delays, cancellations, diver… view at source ↗

**Figure 4.** Figure 4: Class balance between on-time (blue) and delayed (red) departures [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

**Figure 5.** Figure 5: Operational correlation in real (blue) vs. synthetic (red) diversions. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

**Figure 8.** Figure 8: Fidelity evaluation (lower is better). 4.6. Utility Assessment The multi-objective tuning process was designed to maximise the PR-AUC of the classifier predicting flight diversions. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗

**Figure 7.** Figure 7: Statistical evaluation (higher is better). [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗

**Figure 10.** Figure 10: Impact of augmentation size on utility scores. [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗

read the original abstract

Flight diversions are rare but high-impact events in aviation, making their reliable prediction vital for both safety and operational efficiency. However, their scarcity in historical records impedes the training of machine learning models utilised to predict them. This study addresses this scarcity gap by investigating how generative models can augment historical flight data with synthetic diversion records to enhance model training and improve predictive accuracy. We propose a multi-objective optimisation framework coupled with automated hyperparameter search to identify optimal configurations for three deep generative models: Tabular Variational Autoencoder (TVAE), Conditional Tabular Generative Adversarial Network (CTGAN), and CopulaGAN, with the Gaussian Copula (GC) model serving as a statistical baseline. The quality of the synthetic data was examined through a six-stage evaluation framework encompassing realism, diversity, operational validity, statistical similarity, fidelity, and predictive utility. Results show that the optimised models significantly outperform their non-optimised counterparts, and that synthetic augmentation substantially improves diversion prediction compared to models trained solely on real data. These findings demonstrate the effectiveness of hyperparameter-optimised generative models for advancing predictive modelling of rare events in air transportation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper applies multi-objective optimization to tune generative models for rare flight diversion data and reports better predictions, but the search may leak test information.

read the letter

The core takeaway is that optimizing TVAE, CTGAN, and CopulaGAN via a multi-objective framework that includes predictive utility produces synthetic diversion records that improve downstream classifiers over real-data-only training. They also beat the non-optimized versions of the same generators and include a Gaussian Copula baseline plus a six-stage check for realism, diversity, fidelity, and operational validity. This is a targeted engineering application to a concrete aviation imbalance problem rather than a new algorithm.

Referee Report

1 major / 2 minor

Summary. The manuscript proposes a multi-objective optimisation framework with automated hyperparameter search to tune three deep generative models (TVAE, CTGAN, CopulaGAN) plus a Gaussian Copula baseline for producing synthetic flight diversion records. These records are used to augment imbalanced historical flight data, with quality assessed via a six-stage evaluation covering realism, diversity, operational validity, statistical similarity, fidelity and predictive utility. The central empirical claim is that the optimised generators outperform their non-optimised versions and that models trained on the augmented data achieve substantially better diversion prediction than models trained on real data alone.

Significance. If the reported gains are shown to arise from unbiased, realistic synthetic records rather than search artefacts, the work would offer a practical template for rare-event augmentation in safety-critical tabular domains. The explicit multi-objective treatment of predictive utility alongside statistical fidelity is a methodological strength that could be adopted more widely.

major comments (1)

The description of the hyperparameter search (Methods and Evaluation sections) states that predictive utility is one of the objectives optimised for TVAE, CTGAN and CopulaGAN. It is not stated whether the downstream classifier used to compute this utility is trained and evaluated on a validation split that is strictly disjoint from the final held-out test flights. If any part of the search or scalarisation uses the test distribution, the comparison between augmented and real-only training becomes circular and the claimed improvement cannot be interpreted as evidence of realistic augmentation. Please provide the exact data-partitioning diagram or pseudocode and confirm that test data never enters the generator optimisation loop.

minor comments (2)

The abstract asserts 'significant' outperformance without reporting any numerical metrics (AUC, F1, precision-recall, etc.). Adding at least the headline deltas and the number of runs would make the claim immediately verifiable.
Notation for the multi-objective scalarisation (e.g., weights or Pareto-front selection rule) is introduced without an equation number; adding an explicit formulation would aid reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback on our manuscript. The concern regarding data partitioning and potential leakage in the hyperparameter search is well taken, and we address it directly below with a commitment to revise the manuscript for full transparency.

read point-by-point responses

Referee: The description of the hyperparameter search (Methods and Evaluation sections) states that predictive utility is one of the objectives optimised for TVAE, CTGAN and CopulaGAN. It is not stated whether the downstream classifier used to compute this utility is trained and evaluated on a validation split that is strictly disjoint from the final held-out test flights. If any part of the search or scalarisation uses the test distribution, the comparison between augmented and real-only training becomes circular and the claimed improvement cannot be interpreted as evidence of realistic augmentation. Please provide the exact data-partitioning diagram or pseudocode and confirm that test data never enters the generator optimisation loop.

Authors: We agree that the current manuscript does not explicitly describe the data partitioning used during hyperparameter search and predictive-utility evaluation. In the revised version we will add both a data-partitioning diagram (Figure X) and pseudocode (Algorithm Y) in the Methods section. The original flight records are partitioned once, before any model training, into a training set (70 %), a validation set (15 %), and a strictly held-out test set (15 %). All generator hyperparameter optimisation—including the multi-objective search that incorporates predictive utility—operates exclusively on the training and validation portions. The downstream classifier used to compute predictive utility is trained on the training split and evaluated on the validation split; its performance never influences the final test-set evaluation. The test set is used only once, after all generator and classifier training is complete, to measure the final diversion-prediction performance of models trained on real-only versus augmented data. Consequently, test data never enters the generator optimisation loop, and the reported gains are not circular. We will also state this explicitly in the text. revision: yes

Circularity Check

0 steps flagged

No significant circularity: empirical evaluation on held-out real data remains independent of generator optimization

full rationale

The paper describes a multi-objective hyperparameter search for generative models (TVAE, CTGAN, CopulaGAN) that includes predictive utility among six evaluation stages, followed by training a downstream diversion classifier on the resulting augmented data and measuring performance on held-out real flights. No equations, definitions, or self-citations reduce the reported performance gains to quantities that are defined by the same fitted parameters or by the search objective itself. Standard ML practice of separate validation for generator tuning and a final independent test set keeps the central claim (augmentation improves real-data prediction) self-contained and externally falsifiable.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the premise that generative models can faithfully augment rare tabular events and that the chosen evaluation metrics correctly measure predictive utility; many model hyperparameters are treated as searchable rather than fixed a priori.

free parameters (1)

Hyperparameters of TVAE, CTGAN, and CopulaGAN
Automatically searched within the multi-objective optimization framework; their final values are fitted to the flight data.

axioms (1)

domain assumption Generative models trained on imbalanced tabular flight records can produce synthetic diversion samples that preserve statistical and operational properties of real diversions
Invoked throughout the augmentation and evaluation stages.

pith-pipeline@v0.9.0 · 5509 in / 1167 out tokens · 40098 ms · 2026-05-10T00:42:41.852716+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

208 extracted references · 74 canonical work pages

[1]

2020 , url =

Automation in Air Traffic Management: Long-term Vision and Initial Research Roadmap , institution =. 2020 , url =

2020
[2]

2020 , url =

European ATM Master Plan - Digitalising Europe's Aviation Infrastructure , institution =. 2020 , url =

2020
[3]

2020 , url =

Fly AI Report - Demystifying and Accelerating AI in Aviation/ATM , institution =. 2020 , url =

2020
[4]

2023 , url =

SESAR Innovation Pipeline - Air Traffic Management Research and Innovation , institution =. 2023 , url =

2023
[5]

2020 , url =

The Impact of the General Data Protection Regulation (GDPR) on Artificial Intelligence , institution =. 2020 , url =

2020
[6]

2016 , url =

General Data Protection Regulation (GDPR) , author =. 2016 , url =

2016
[7]

2022 , month =

Automation in Air Traffic Control: Trust, Teamwork, Resilience, Safety , journal =. 2022 , month =. doi:10.1016/j.trpro.2022.11.003 , url =

work page doi:10.1016/j.trpro.2022.11.003 2022
[8]

Machine Learning for Synthetic Data Generation: A Review , journal =

Yingzhou Lu and Minjie Shen and Huazheng Wang and Xiao Wang and Capucine van Rechem and Wenqi Wei , year =. Machine Learning for Synthetic Data Generation: A Review , journal =
[9]

A Survey on Data Collection for Machine Learning: A Big Data - AI Integration Perspective , year=

Roh, Yuji and Heo, Geon and Whang, Steven Euijong , journal=. A Survey on Data Collection for Machine Learning: A Big Data - AI Integration Perspective , year=. doi:10.1109/TKDE.2019.2946162 , url=

work page doi:10.1109/tkde.2019.2946162 2019
[10]

and Dervovic, Danial and Mahfouz, Mahmoud and Tillman, Robert E

Assefa, Samuel A. and Dervovic, Danial and Mahfouz, Mahmoud and Tillman, Robert E. and Reddy, Prashant and Veloso, Manuela , title =. 2021 , isbn =. doi:10.1145/3383455.3422554 , booktitle =

work page doi:10.1145/3383455.3422554 2021
[11]

PLOS Digital Health , publisher =

Synthetic data in health care: A narrative review , year =. PLOS Digital Health , publisher =. doi:10.1371/journal.pdig.0000082 , author =

work page doi:10.1371/journal.pdig.0000082
[12]

Synthetic to Real Gap Estimation of Autonomous Driving Datasets using Feature Embedding , year=

Gadipudi, Nivesh and Elamvazuthi, Irraivan and Sanmugam, Mahindra and Izhar, Lila Iznita and Prasetyo, Tindyo and Jegadeeshwaran, R and Ali, Syed Saad Azhar , booktitle=. Synthetic to Real Gap Estimation of Autonomous Driving Datasets using Feature Embedding , year=. doi:10.1109/ROMA55875.2022.9915679 , url=

work page doi:10.1109/roma55875.2022.9915679 2022
[13]

Generative Adversarial Nets , url =

Goodfellow, Ian and Pouget-Abadie, Jean and Mirza, Mehdi and Xu, Bing and Warde-Farley, David and Ozair, Sherjil and Courville, Aaron and Bengio, Yoshua , journal =. Generative Adversarial Nets , url =. 2014 , month =

2014
[14]

Kingma and Max Welling

Kingma, Diederik P. and Welling, Max , title =. 2019 , volume =. doi:10.1561/2200000056 , journal =

work page doi:10.1561/2200000056 2019
[15]

Generative adversarial networks,

Goodfellow, Ian and Pouget-Abadie, Jean and Mirza, Mehdi and Xu, Bing and Warde-Farley, David and Ozair, Sherjil and Courville, Aaron and Bengio, Yoshua , title =. 2020 , issue_date =. doi:10.1145/3422622 , journal =

work page doi:10.1145/3422622 2020
[16]

Denoising Diffusion Probabilistic Models , url =

Ho, Jonathan and Jain, Ajay and Abbeel, Pieter , booktitle =. Denoising Diffusion Probabilistic Models , url =
[17]

Tackling the Generative Learning Trilemma with Denoising Diffusion GANs , journal =

Zhisheng Xiao and Karsten Kreis and Arash Vahdat , year =. Tackling the Generative Learning Trilemma with Denoising Diffusion GANs , journal =
[18]

Cohen and Adrian Weller , year =

James Jordon and Lukasz Szpruch and Florimond Houssiau and Mirko Bottarelli and Giovanni Cherubin and Carsten Maple and Samuel N. Cohen and Adrian Weller , year =. Synthetic Data -- what, why and how? , journal =
[19]

Time-series Generative Adversarial Networks , url =

Yoon, Jinsung and Jarrett, Daniel and van der Schaar, Mihaela , booktitle =. Time-series Generative Adversarial Networks , url =
[20]

2022 , isbn =

Li, Xiaomin and Metsis, Vangelis and Wang, Huangyingrui and Ngu, Anne Hee Hiong , title =. 2022 , isbn =. doi:10.1007/978-3-031-09342-5_13 , booktitle =

work page doi:10.1007/978-3-031-09342-5_13 2022
[21]

and Lee, Yang W

Pipino, Leo L. and Lee, Yang W. and Wang, Richard Y. , title =. 2002 , volume =. doi:10.1145/505248.506010 , journal =

work page doi:10.1145/505248.506010 2002
[22]

2020 , month =

Dina Kampouraki and Robert Riemann and Vìtor Bernardo and Maria Enescu and Marek Jessen and Kostantina Vemou and Stefano Leucci , title =. 2020 , month =

2020
[23]

The Real Deal about Synthetic Data , journal =

Fernando Lucini , year =. The Real Deal about Synthetic Data , journal =
[24]

Proceedings of the 11th International Conference on Information Quality, MIT, Cambridge, MA, USA, November 10-12, 2006 , pages =

Valerie Sessions and Marco Valtorta , title =. Proceedings of the 11th International Conference on Information Quality, MIT, Cambridge, MA, USA, November 10-12, 2006 , pages =. 2006 , url =

2006
[25]

Data scarcity, robustness and extreme multi-label classification , journal =

Rohit Babbar and Bernard Schölkopf , year =. Data scarcity, robustness and extreme multi-label classification , journal =
[26]

Realistic Synthetic Financial Transactions for Anti-Money Laundering Models , url =

Altman, Erik and Blanu. Realistic Synthetic Financial Transactions for Anti-Money Laundering Models , url =. Advances in Neural Information Processing Systems , pages =
[27]

2022 , isbn =

Liu, Chunli and Ventre, Carmine and Polukarov, Maria , title =. 2022 , isbn =. doi:10.1145/3533271.3561704 , booktitle =

work page doi:10.1145/3533271.3561704 2022
[28]

2010 , issn =

Using micro-simulation to create a synthesised data set and test policy options: The case of health service effects under demographic ageing , journal =. 2010 , issn =. doi:10.1016/j.healthpol.2010.05.014 , url =

work page doi:10.1016/j.healthpol.2010.05.014 2010
[29]

2021 , isbn =

Sun, Siao and Wang, Fusheng and Rashidian, Sina and Kurc, Tahsin and Abell-Hart, Kayley and Hajagos, Janos and Zhu, Wei and Saltz, Mary and Saltz, Joel , title =. 2021 , isbn =. doi:10.1007/978-3-030-93663-1_12 , pages =

work page doi:10.1007/978-3-030-93663-1_12 2021
[30]

Virtual passengers for real car solutions: synthetic datasets , journal =

Paola Natalia Canas and Juan Diego Ortega and Marcos Nieto and Oihana Otaegui , year =. Virtual passengers for real car solutions: synthetic datasets , journal =
[31]

2024 , issn =

Fake it till you make it: Synthetic data for emerging carsharing programs , journal =. 2024 , issn =. doi:10.1016/j.trd.2024.104067 , url =

work page doi:10.1016/j.trd.2024.104067 2024
[32]

2022 , issn =

A Novel Method in Intelligent Synthetic Data Creation for Machine Learning-based Manufacturing Quality Control , journal =. 2022 , issn =. doi:10.1016/j.ifacol.2022.09.186 , url =

work page doi:10.1016/j.ifacol.2022.09.186 2022
[33]

2024 , issn =

A comparison of methods for generating synthetic training data for domain adaption of deep learning models in ultrasonic non-destructive evaluation , journal =. 2024 , issn =. doi:10.1016/j.ndteint.2023.102978 , url =

work page doi:10.1016/j.ndteint.2023.102978 2024
[34]

2024 , month =

D.2.1 State of the art , institution =. 2024 , month =

2024
[35]

Unified Training of Universal Time Series Forecasting Transformers , journal =

Gerald Woo and Chenghao Liu and Akshat Kumar and Caiming Xiong and Silvio Savarese and Doyen Sahoo , year =. Unified Training of Universal Time Series Forecasting Transformers , journal =
[36]

2020 , issn =

Daniil Ryabko , title =. 2020 , issn =. doi:10.1007/978-3-030-54304-4 , url =

work page doi:10.1007/978-3-030-54304-4 2020
[37]

2024 , issn =

Can I trust my fake data – A comprehensive quality assessment framework for synthetic tabular data in healthcare , journal =. 2024 , issn =. doi:10.1016/j.ijmedinf.2024.105413 , url =

work page doi:10.1016/j.ijmedinf.2024.105413 2024
[38]

2022 , issn =

Application of tabular data synthesis using generative adversarial networks on machine learning-based multiaxial fatigue life prediction , journal =. 2022 , issn =. doi:10.1016/j.ijpvp.2022.104779 , url =

work page doi:10.1016/j.ijpvp.2022.104779 2022
[39]

2024 , issn =

ConvGeN: A convex space learning approach for deep-generative oversampling and imbalanced classification of small tabular datasets , journal =. 2024 , issn =. doi:10.1016/j.patcog.2023.110138 , url =

work page doi:10.1016/j.patcog.2023.110138 2024
[40]

Synthetic data generation for tabular health records: A systematic review

Synthetic data generation for tabular health records: A systematic review , journal =. 2022 , issn =. doi:10.1016/j.neucom.2022.04.053 , url =

work page doi:10.1016/j.neucom.2022.04.053 2022
[41]

2024 , series =

Chapter 8 - Medical image synthesis and reconstruction using generative adversarial networks , booktitle =. 2024 , series =. doi:10.1016/B978-0-32-385124-4.00018-0 , url =

work page doi:10.1016/b978-0-32-385124-4.00018-0 2024
[42]

2023 , month =

Synthetic image data generation using BIM and computer graphics for building scene understanding , journal =. 2023 , month =. doi:10.1016/j.autcon.2023.105016 , url =

work page doi:10.1016/j.autcon.2023.105016 2023
[43]

2024 , month =

A survey of GPT-3 family large language models including ChatGPT and GPT-4 , journal =. 2024 , month =. doi:10.1016/j.nlp.2023.100048 , url =

work page doi:10.1016/j.nlp.2023.100048 2024
[44]

A survey on text generation using generative adversarial networks , journal =

Gustavo H. A survey on text generation using generative adversarial networks , journal =. 2021 , month =. doi:10.1016/j.patcog.2021.108098 , url =

work page doi:10.1016/j.patcog.2021.108098 2021
[45]

2023 , month =

Generative emotional AI for speech emotion recognition: The case for synthetic emotional speech augmentation , journal =. 2023 , month =. doi:10.1016/j.apacoust.2023.109425 , url =

work page doi:10.1016/j.apacoust.2023.109425 2023
[46]

2022 , month =

GAN computers generate arts? A survey on visual arts, music, and literary text generation using generative adversarial network , journal =. 2022 , month =. doi:10.1016/j.displa.2022.102237 , url =

work page doi:10.1016/j.displa.2022.102237 2022
[47]

2022 , month =

Synthetic data generation with deep generative models to enhance predictive tasks in trading strategies , journal =. 2022 , month =. doi:https://doi.org/10.1016/j.ribaf.2022.101747 , url =

work page doi:10.1016/j.ribaf.2022.101747 2022
[48]

S., Kakati, P., Dandotiya, D., M, U

Synthetic demand data generation for individual electricity consumers : Generative Adversarial Networks (GANs) , journal =. 2022 , month =. doi:https://doi.org/10.1016/j.egyai.2022.100161 , url =

work page doi:10.1016/j.egyai.2022.100161 2022
[49]

Sensors , volume =

Hu, Chaochen and Sun, Zihan and Li, Chao and Zhang, Yong and Xing, Chunxiao , title =. Sensors , volume =. 2023 , month =

2023
[50]

2024 , month =

FIND: A Synthetic weather generator to control drought Frequency, Intensity, and Duration , journal =. 2024 , month =. doi:https://doi.org/10.1016/j.envsoft.2023.105927 , url =

work page doi:10.1016/j.envsoft.2023.105927 2024
[51]

2023 , month =

A spatially-dependent synthetic global dataset of extreme sea level events , journal =. 2023 , month =. doi:10.1016/j.wace.2023.100596 , url =

work page doi:10.1016/j.wace.2023.100596 2023
[52]

SynMob: Creating High-Fidelity Synthetic GPS Trajectory Dataset for Urban Mobility Analysis , url =

Zhu, Yuanshao and Ye, Yongchao and Wu, Ying and Zhao, Xiangyu and Yu, James , booktitle =. SynMob: Creating High-Fidelity Synthetic GPS Trajectory Dataset for Urban Mobility Analysis , url =
[53]

2021 , month =

Generating a synthetic probabilistic daily activity-location schedule using large-scale, long-term and low-frequency smartphone GPS data with limited activity information , journal =. 2021 , month =. doi:10.1016/j.trc.2021.103408 , url =

work page doi:10.1016/j.trc.2021.103408 2021
[54]

Artificial Intelligence in Education

Leiker, Daniel and Gyllen, Ashley Ricker and Eldesouky, Ismail and Cukurova, Mutlu , title=. Artificial Intelligence in Education. Posters and Late Breaking Results, Workshops and Tutorials, Industry and Innovation Tracks, Practitioners, Doctoral Consortium and Blue Sky , year=
[55]

A Simple Text to Video Model via Transformer , journal =

Gang Chen , year =. A Simple Text to Video Model via Transformer , journal =
[56]

Proceedings of the Fifth International Scientific Conference ``Intelligent Information Technologies for Industry'' (IITI'21) , year=

Polyakov, Vladimir and Mezhenin, Aleksandr and Vereskun, Vladimir and Korobeynikov, Anatoly and Shapovalov, Vasilii , title=. Proceedings of the Fifth International Scientific Conference ``Intelligent Information Technologies for Industry'' (IITI'21) , year=
[57]

Bringing up OpenSky: a large-scale ADS-B sensor network for research , year =

Sch\". Bringing up OpenSky: a large-scale ADS-B sensor network for research , year =. Proceedings of the 13th International Symposium on Information Processing in Sensor Networks , pages =
[58]

Aviation Data for Research , author =
[59]

Journal of Open Source Software , title=

Xavier. Journal of Open Source Software , title=. 2019 , volume=. doi:10.21105/joss.01518 , issn=

work page doi:10.21105/joss.01518 2019
[60]

2023 , publisher=

Generative Deep Learning: Teaching Machines to Paint, Write, Compose, and Play , author=. 2023 , publisher=

2023
[61]

Improving Aircraft Turnaround Reliability , booktitle =

Schultz, Michael and Fricke, Hartmut , year =. Improving Aircraft Turnaround Reliability , booktitle =
[62]

In: 2021 IEEE/AIAA 40th Digital Avionics Systems Conference (DASC)

Luo, Mingchuan and Schultz, Michael and Fricke, Hartmut and Desart, Bruno and Herrema, Floris and Barragán Montes, Rocío , booktitle=. Agent-based simulation for aircraft stand operations to predict ground time using machine learning , year=. doi:10.1109/DASC52595.2021.9594325 , url=

work page doi:10.1109/dasc52595.2021.9594325 2021
[63]

Operations Research Proceedings 2019 , year=

Asadi, Ehsan and Evler, Jan and Preis, Henning and Fricke, Hartmut , title=. Operations Research Proceedings 2019 , year=

2019
[64]

and Brandão, Diego and Carvalho, Diego and Ogasawara, Eduardo , year =

Carvalho, Leonardo and Sternberg, Alice and Maia Gonçalves, Leandro and Beatriz Cruz, Ana and Soares, Jorge A. and Brandão, Diego and Carvalho, Diego and Ogasawara, Eduardo , year =. On the relevance of data science for flight delay research: a systematic review , journal =
[65]

A Review of Flight Delay Prediction Methods , year=

Wang, Teng and Zheng, Yufan and Xu, Haiwen , booktitle=. A Review of Flight Delay Prediction Methods , year=. doi:10.1109/BDEE55929.2022.00029 , url=

work page doi:10.1109/bdee55929.2022.00029 2022
[66]

2015 , month =

Endogenous control of service rates in stochastic and dynamic queuing models of airport congestion , journal =. 2015 , month =. doi:10.1016/j.tre.2014.10.014 , url =

work page doi:10.1016/j.tre.2014.10.014 2015
[67]

Agent-based vulnerability assessment at airport security checkpoints: A case study on security operator behavior , journal =

Stef Janssen and Arjan. Agent-based vulnerability assessment at airport security checkpoints: A case study on security operator behavior , journal =. 2020 , month =. doi:10.1016/j.trip.2020.100139 , url =

work page doi:10.1016/j.trip.2020.100139 2020
[68]

Aerospace , volume =

Janssen, Stef and van der Sommen, Régis and Dilweg, Alexander and Sharpanskykh, Alexei , title =. Aerospace , volume =. 2020 , number =

2020
[69]

Applied Sciences , volume =

Chen, Bin and Zhao, Xing and Wu, Jin , title =. Applied Sciences , volume =. 2023 , number =

2023
[70]

Validation of Simulated Synthetic Air Traffic against Recorded Air Traffic in Germany , url =

Thomas Gräupl , booktitle=. Validation of Simulated Synthetic Air Traffic against Recorded Air Traffic in Germany , url =. 2017 , month =

2017
[71]

2023 , month =

Deep generative modelling of aircraft trajectories in terminal maneuvering areas , journal =. 2023 , month =. doi:10.1016/j.mlwa.2022.100446 , url =

work page doi:10.1016/j.mlwa.2022.100446 2023
[72]

2023 , publisher =

Dalmau, Ramon and Gawinowski, Gilles , title =. 2023 , publisher =. doi:10.1109/TITS.2023.3235741 , journal =

work page doi:10.1109/tits.2023.3235741 2023
[73]

2024 , month =

The effectiveness of supervised clustering for characterising flight diversions due to weather , journal =. 2024 , month =. doi:10.1016/j.eswa.2023.121652 , url =

work page doi:10.1016/j.eswa.2023.121652 2024
[74]

Detecting flight trajectory anomalies and predicting diversions in freight transportation , journal =

Claudio. Detecting flight trajectory anomalies and predicting diversions in freight transportation , journal =. 2016 , month =. doi:10.1016/j.dss.2016.05.004 , url =

work page doi:10.1016/j.dss.2016.05.004 2016
[75]

2018 , url =

SESAR Joint Undertaking , title =. 2018 , url =

2018
[76]

2024 , month =

Discriminative multimodal learning via conditional priors in generative models , journal =. 2024 , month =. doi:10.1016/j.neunet.2023.10.048 , url =

work page doi:10.1016/j.neunet.2023.10.048 2024
[77]

2024 , month =

AI-supported estimation of safety critical wind shear-induced aircraft go-around events utilizing pilot reports , journal =. 2024 , month =. doi:10.1016/j.heliyon.2024.e28569 , url =

work page doi:10.1016/j.heliyon.2024.e28569 2024
[78]

2008 , month =

Runway sequencing with holding patterns , journal =. 2008 , month =. doi:10.1016/j.ejor.2006.06.076 , url =

work page doi:10.1016/j.ejor.2006.06.076 2008
[79]

Year of Publication , note =
[80]

Deep Learning , author=

Showing first 80 references.