Advanced Predictive Analytics for Aircraft Accident Severity Using Deep Learning

S. Thenmalar, B. Jaya Krishna Yadav, D. Venkat Kishore

2025

Abstract

Aviation safety is seriously threatened by aircraft accidents, which calls for sophisticated prediction models for precise severity categorization and risk reduction. The intricate, nonlinear linkages found in accident data are frequently missed by traditional approaches, resulting in less-than-ideal forecasts and postponed preventive actions. Our research uses machine learning models and deep learning techniques to create a sophisticated forecasting system for classifying the severity of plane accidents. To increase the dataset's prediction capacity, we use feature engineering approaches and conduct in-depth Exploratory Data Analysis (EDA) on historical accident data. We apply the XGB-Classifier after thorough processing and data organizing, and it achieves an impressive train accuracy of 100% to evaluate accuracy of 95.9%. We create a model of Convolutional Neural Networks (to improve performance even further, and it first achieves an accurate training of 97.66% and an accuracy in tests of 93.6%. The model's accuracy is enhanced for both low-severity incidents (train: 99.13%, test: 96.17%) and high-severity accidents (train: 99.53%, test: 96.93%) by hyperparameter tuning and severity-specific optimization. By combining both severity levels, the final CNN model shows a strong predictive performance with an improved train precision of 98.30% and test accuracy of 97.93%. These results demonstrate how well-structured preprocessing, feature engineering, and sophisticated deep learning architectures work together to produce a potent tool for immediate accident severity assessment and aviation safety improvement.

Paper Citation

in Harvard Style

Thenmalar S., Yadav B. and Kishore D. (2025). Advanced Predictive Analytics for Aircraft Accident Severity Using Deep Learning. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 648-656. DOI: 10.5220/0013903400004919

in Bibtex Style

@conference{icrdicct`2525,
author={S. Thenmalar and B. Yadav and D. Kishore},
title={Advanced Predictive Analytics for Aircraft Accident Severity Using Deep Learning},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25},
year={2025},
pages={648-656},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013903400004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25
TI - Advanced Predictive Analytics for Aircraft Accident Severity Using Deep Learning
SN - 978-989-758-777-1
AU - Thenmalar S.
AU - Yadav B.
AU - Kishore D.
PY - 2025
SP - 648
EP - 656
DO - 10.5220/0013903400004919
PB - SciTePress