Global Cabin Safety Incident Reports from 2004–2024

8,280 global cabin abnormal event reports from 2004–2024 were used to train a hybrid CNN-LSTM-Attention model for intelligent risk assessment. The model achieved 95.01% accuracy and a 94.17% F1 score on a test set, outperforming benchmark approaches. The dataset, shared by Lianbin Zhou on figshare, provides a foundation for data-driven decision-making in civil aviation safety management.

Use Cases

• Classifying cabin safety incidents based on textual reports using deep learning models.
• Quantifying risk levels by mapping text features to hazard mechanisms.
• Benchmarking machine learning models for performance on aviation safety text classification.
• Supporting data-driven decision-making for civil aviation safety management.

Strengths

• Contains 8,280 incident reports spanning a 20-year period (2004–2024).
• The associated model achieved a reported 95.01% accuracy and 94.17% F1 score, suggesting a well-defined classification task.
• The framework establishes a mapping from text features to risk levels, aiming for interpretable risk quantification.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small at 9.5 KB, indicating limited scope.

Provenance

Source

figshare

Collection Method

Likely compiled from global cabin abnormal event reports.

Time Range

2004–2024

Freshness

Last updated 2026-05-27 17:39:55; freshness should be verified.

Geography

Global