Aviation Incident Classification Dataset with 5-Fold Cross-Validation Configuration

Sawera Qureshi's dataset contains a 5-fold cross-validation configuration for classifying aviation safety occurrence reports. The underlying data comprises 53,770 aviation occurrence summaries from the Transportation Safety Board of Canada, spanning an 80-year period. The dataset was last updated on 2026-05-21.

Use Cases

• Training and evaluating text classifiers for aviation incident/accident reports based on occurrence summaries.
• Benchmarking machine learning models like SVM, Random Forest, and Naive Bayes on a large historical safety dataset.
• Supporting automated triage of safety reports to identify potentially mislabeled cases for expert review.
• Integrating classification outputs into downstream severity assessment pipelines for aviation safety operations.

Strengths

• Dataset is based on 53,770 aviation occurrence summaries, providing a substantial text corpus.
• Data covers an 80-year temporal range, offering a long-term historical perspective.
• The described evaluation framework achieved a reported classification accuracy of 98.06% using SVM.

Limitations

• Row count for the cross-validation configuration is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small at 5.5 KB, indicating it contains only configuration or metadata, not the full text data.

Provenance

Source

Transportation Safety Board of Canada

Collection Method

Summaries of aviation occurrences were collected and processed.

Time Range

Covers an 80-year period (specific years not provided).

Freshness

Last updated 2026-05-21 17:44:30; freshness should be verified.