Asghar, Taha, Mufti, Bilal, Zaheer, Syed Qasim, Ahmed, Luqman · Nov 5, 2026
Taha Asghar and coauthors: Hybrid Machine Learning for Imbalanced Landing-Phase Safety Event Prediction from Flight Data
Markdown · JSON · Original source
The paper’s authors propose an interpretable hybrid feature-fusion framework to identify rare abnormal approaches in flight data. Their method combines an LSTM autoencoder’s reconstruction-error descriptor with statistical flight-history features, then uses a class-weighted XGBoost classifier to distinguish four types of landing-phase window: Normal, Speed High, Path High, and Flaps Late. The authors evaluate it on 53,750 DASHlink approach windows from three arrival airports. They report mean macro recall above 0.90 across all airports under repeated partitioning and retraining. In their comparisons, continuous feature fusion outperforms reconstruction-error-only and hard-gate alternatives; airport-specific training benefits the two airports with larger datasets, while pooled training benefits the airport with the smallest dataset. The authors also report that feature attributions associate predictions with vertical-path, configuration, speed, and approach-context variables. The work’s practical point is that data availability may affect whether flight-data monitoring models should be trained for a particular airport or pooled across airports. Escape Meridian’s reading: these results concern the evaluated dataset and modeling comparisons; the abstract does not report operational deployment.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Nov 5, 2026 (2026-11-05).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-08T17:38:18.570Z. First observed: 2026-11-06T05:10:34.692Z. Retrieved: 2026-11-08T17:33:15.762Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The evaluation described in the abstract uses DASHlink approach windows from three arrival airports; it does not establish whether the reported results generalize to other airports. The abstract reports modeling results and feature attributions, but does not report operational deployment.
Open question
Would the reported airport-specific and pooled-training results hold at airports beyond the three evaluated?
Source-stated
The authors propose an interpretable hybrid feature-fusion framework for multiclass landing-phase safety-event prediction.
This study proposes an interpretable hybrid feature-fusion framework for multiclass landing-phase safety-event prediction.
Source-stated
The framework combines an LSTM autoencoder’s reconstruction-error descriptor with statistical flight-history features and supplies them to a class-weighted XGBoost classifier.
Its continuous reconstruction-error descriptor is fused with parameter-specific statistical flight-history features and supplied to a class-weighted XGBoost classifier
Source-stated
The evaluation uses 53,750 NASA DASHlink approach windows from three arrival airports.
The framework is evaluated using 53,750 NASA DASHlink approach windows from three arrival airports.
Source-stated
The authors report mean macro recall above 0.90 across all airports under repeated partitioning and retraining.
with mean macro recall above 0.90 across all airports under repeated partitioning and retraining.
Source-stated
The authors report that local specialization benefits the two airports with larger datasets, while pooled training benefits the airport with the smallest dataset.
Airport-level comparisons reveal benefits from local specialization at the two airports with larger datasets and from pooled training at the airport with the smallest dataset.
Escape Meridian interpretation
The reported results concern the evaluated dataset and modeling comparisons; the abstract does not report operational deployment.
The framework is evaluated using 53,750 NASA DASHlink approach windows from three arrival airports.
Source and contributor context
Asghar, Taha, Mufti, Bilal, Zaheer, Syed Qasim, Ahmed, Luqman · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Machine learning
Methods that let computers learn patterns from examples instead of following rules written by hand.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
NASA
NASA is the US agency for civilian space exploration, aeronautics and related scientific research.
Editorial provenance
Version 1. Publication ID: 17a1214e-4b8f-43e5-855f-83578ddb440c.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T17:38:18.570Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: b11d2da6d6e8221523b6ec4a697e02aba39489a831714b0dfebac50b32773a03. Source text hash: 173b7e72f8ab7b0e3526845dbe8d240661d87778da4bc0d0483aa522a005abbf.
Revision and source-check history
Version 1 · 2026-11-08T17:38:18.570Z
Taha Asghar and coauthors: Hybrid Machine Learning for Imbalanced Landing-Phase Safety Event Prediction from Flight Data
The paper’s authors propose an interpretable hybrid feature-fusion framework to identify rare abnormal approaches in flight data. Their method combines an LSTM autoencoder’s reconstruction-error descriptor with statistical flight-history features, then uses a class-weighted XGBoost classifier to distinguish four types of landing-phase window: Normal, Speed High, Path High, and Flaps Late. The authors evaluate it on 53,750 DASHlink approach windows from three arrival airports. They report mean macro recall above 0.90 across all airports under repeated partitioning and retraining. In their comparisons, continuous feature fusion outperforms reconstruction-error-only and hard-gate alternatives; airport-specific training benefits the two airports with larger datasets, while pooled training benefits the airport with the smallest dataset. The authors also report that feature attributions associate predictions with vertical-path, configuration, speed, and approach-context variables. The work’s practical point is that data availability may affect whether flight-data monitoring models should be trained for a particular airport or pooled across airports. Escape Meridian’s reading: these results concern the evaluated dataset and modeling comparisons; the abstract does not report operational deployment.
Original source
cs.AI · Article
Source details
Original source date: Nov 5, 2026 (2026-11-05).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-08T17:38:18.570Z. First observed: 2026-11-06T05:10:34.692Z. Retrieved: 2026-11-08T17:33:15.762Z.
Limits
The evaluation described in the abstract uses DASHlink approach windows from three arrival airports; it does not establish whether the reported results generalize to other airports. The abstract reports modeling results and feature attributions, but does not report operational deployment.
Open question
Would the reported airport-specific and pooled-training results hold at airports beyond the three evaluated?
Source-stated
The authors propose an interpretable hybrid feature-fusion framework for multiclass landing-phase safety-event prediction.
This study proposes an interpretable hybrid feature-fusion framework for multiclass landing-phase safety-event prediction.
Source-stated
The framework combines an LSTM autoencoder’s reconstruction-error descriptor with statistical flight-history features and supplies them to a class-weighted XGBoost classifier.
Its continuous reconstruction-error descriptor is fused with parameter-specific statistical flight-history features and supplied to a class-weighted XGBoost classifier
Source-stated
The evaluation uses 53,750 NASA DASHlink approach windows from three arrival airports.
The framework is evaluated using 53,750 NASA DASHlink approach windows from three arrival airports.
Source-stated
The authors report mean macro recall above 0.90 across all airports under repeated partitioning and retraining.
with mean macro recall above 0.90 across all airports under repeated partitioning and retraining.
Source-stated
The authors report that local specialization benefits the two airports with larger datasets, while pooled training benefits the airport with the smallest dataset.
Airport-level comparisons reveal benefits from local specialization at the two airports with larger datasets and from pooled training at the airport with the smallest dataset.
Escape Meridian interpretation
The reported results concern the evaluated dataset and modeling comparisons; the abstract does not report operational deployment.
The framework is evaluated using 53,750 NASA DASHlink approach windows from three arrival airports.
Source and contributor context
Asghar, Taha, Mufti, Bilal, Zaheer, Syed Qasim, Ahmed, Luqman · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Machine learning
Methods that let computers learn patterns from examples instead of following rules written by hand.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
NASA
NASA is the US agency for civilian space exploration, aeronautics and related scientific research.
Editorial provenance
Version 1. Publication ID: 17a1214e-4b8f-43e5-855f-83578ddb440c.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T17:38:18.570Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: b11d2da6d6e8221523b6ec4a697e02aba39489a831714b0dfebac50b32773a03. Source text hash: 173b7e72f8ab7b0e3526845dbe8d240661d87778da4bc0d0483aa522a005abbf.
uncertain · 2026-11-08T17:38:18.570Z · anthropic/claude-opus-5.5
Publication: 17a1214e-4b8f-43e5-855f-83578ddb440c
attribution — pass: The title credits Taha Asghar and coauthors, matching the listed authors. The summary attributes the proposal, evaluation and results to the authors. The deployment caveat is clearly labeled as Escape Meridian's reading.
Authors:Taha Asghar, Bilal Mufti, Syed Qasim Zaheer, Luqman Ahmed
summary — uncertain: The summary is faithful to the abstract, and the question follows from the three-airport scope. However, central technical terms are left unexplained for non-specialists: LSTM autoencoder, reconstruction-error descriptor, class-weighted XGBoost, hard-gate and especially macro recall. Macro recall matters most because the headline result (above 0.90) cannot be interpreted without knowing it roughly measures how often each event type, including rare ones, is caught. Keep the terms exactly and add term-library tooltips for macro recall, LSTM autoencoder, XGBoost and hard-gate. The summary could also briefly say that windows are segments of flight data from landing approaches.
with mean macro recall above 0.90 across all airports under repeated partitioning and retraining.
date — pass: originalPublished 2026-11-05 matches the v1 submission date. The feed date of 2026-11-06 is a later announcement.
[Submitted on 5 Nov 2026]
support — pass: All claims are directly stated in the abstract with accurate excerpts. The practical-point sentence uses hedged language ('may affect') and is consistent with the stated practical guidance on airport-level data availability. The interpretation claim is labeled and does not overstate. No benefits, costs or real-world reliability are invented, and the test results are not presented as demonstrated operational safety gains.
practical guidance for adapting flight-data monitoring models to airport-level data availability
media — pass: The selected media is the verified direct arXiv abstract page for this paper. It is the most direct way to access the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
NASA is the US agency for civilian space exploration, aeronautics and related scientific research.
Explore NASAarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivNASA is the US agency for civilian space exploration, aeronautics and related scientific research.
Explore NASANASA is the US agency for civilian space exploration, aeronautics and related scientific research.
Explore NASAarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivNASA is the US agency for civilian space exploration, aeronautics and related scientific research.
Explore NASA