Machine learning-based anomaly detection and prediction in commercial aircraft using autonomous surveillance data

Area of Science:

• Aviation Safety
• Machine Learning
• Data Science

Background:

• Commercial aviation is vital for global transport, yet accidents, though rare, pose significant risks.
• Predictive models are needed to enhance safety by detecting and forecasting potential aircraft incidents.

Purpose of the Study:

• To develop and evaluate a machine learning model for detecting and predicting commercial aircraft accidents.
• To assess the model's accuracy in categorizing anomalies and recognizing data patterns.

Main Methods:

• Formulated the problem, selected and labeled data according to Global Aviation Organisation standards.
• Developed and tested a machine learning model, specifically linear dipole testing, for accident prediction.
• Validated data tagging with expert business pilots.

Main Results:

• The linear dipole testing model achieved 93% precision, indicating high accuracy in prediction.
• The model demonstrated strong performance with an area-under-the-curve of 0.97 for anomaly identification.
• An area-under-the-curve of 0.96 was recorded for daily detection, confirming the model's reliability.

Conclusions:

• The developed machine learning model is highly effective for predicting commercial aircraft accidents.
• The model's high precision and AUC scores validate its capability for anomaly and daily detection in aviation.
• This predictive capability can significantly contribute to improving overall aviation safety and operational efficiency.