Data-Driven Decision Making in Aviation Safety Management Systems: A Supervised Machine Learning Approach.

Article Sidebar

pdf
Published: May 8, 2026
Keywords:
Machine Learning Supervised Learning Safety Management Systems (SMS) Proactive and Predictive Safety Radom Forest Support Vector Machine Extreme Gradient Boosting SMOTE Technique

Main Article Content

Vivek Sharma
Florida Institute of Technology
https://orcid.org/0000-0002-8758-8606
Brooke Wheeler
Florida Institute of Technology
https://orcid.org/0000-0001-5507-729X
Bhoomin Chauhan
NASA AMES Research Center
Shaun Kelly
Purdue University
https://orcid.org/0009-0001-6060-3597

Abstract

One of the objectives of the FAA’s safety improvement plan is to continuously collect safety data to identify potential risks. Over the last decade, the application of machine learning (ML) and Artificial Intelligence (AI) models in prediction, classification, and identification has been widely used across both aviation and nonaviation domains. Few studies have explored the use of ML techniques in aviation for predicting safety. Therefore, the purpose of the current study is twofold: (a) to build and compare the classification performance of three supervised machine learning models: Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boost (XG Boost), and (b) to apply the Synthetic Minority Oversampling Technique (SMOTE) class imbalance technique to all three models and compare the performance.  The total sample size for the current study was N = 17,275, of which 15,870 (91.86%) were classified as accidents and 1,405 (9.14%) were classified as incidents. The NTSB database includes many features, such as event ID, registration number, Federal Aviation Regulation (FAR), flight plan, damage type, event type, crew demographics, and aircraft characteristics. The dependent or outcome variable for the current study was event type (accident or incident). The findings of this study demonstrate that supervised ML models can be effectively used to predict or classify aviation events such as incidents or accidents. Specifically, the Random Forest, SVM, and XG Boost models can be applied to operational data to classify aviation accidents and incidents. The findings of this study offer multiple practical implications for enhancing safety through proactive and predictive data analytical decision-making approaches, which align with ICAO’s (2018) SMM. A robust SMS is always dependent on data-driven decision-making, and integrating ML to proactively identify hazards can strengthen the SMS foundation.


 

Article Details

Issue
Vol. 44 No. 1 (2026): Collegiate Aviation Review International
Section
Peer-Reviewed Articles
Author Biographies

Vivek Sharma, Florida Institute of Technology

Dr. Vivek Sharma is an Assistant Professor in the College of Aeronautics (COA) at the Florida Institute of Technology and serves as the COA Division Director and Chair of Online Programs. He holds a Ph.D. in Aviation Sciences with a concentration in Human Factors and Safety. Dr. Sharma has extensive research experience in aviation safety, human factors, aviation data analytics, Unmanned Aircraft Systems (UAS), Advanced and Urban Air Mobility (AAM/UAM), and user experience (UX). Dr. Sharma teaches graduate and undergraduate courses in Aviation Statistics, Human Factors, Multiple Regression and Correlation, Aviation Data Analytics, Aviation Research Methods, Unmanned Aircraft Systems, and Aviation Safety.

His research portfolio includes analyzing operational data generated through Flight Operations Quality Assurance (FOQA)/ Flight Data Monitoring (FDM) programs. Applying supervised and unsupervised machine learning (ML) approaches, including linear and logistic regression, Random Forest, Support Vector Machine (SVM), Gradient Boost (XG Boost), and K-means Clustering, to large-scale aviation datasets to support proactive safety decision-making. He has served as a lead researcher on several projects for the THRUST Aviation Research Lab, ATLAS Lab, and the Federal Aviation Administration’s PEGASAS Center of Excellence, applying advanced statistical methods and Tableau-based data visualization to translate complex aviation datasets into actionable safety insights. Additionally, his focus on Unmanned Aircraft Systems (UAS) research involves addressing human factors and safety considerations for Part 107, recreational, public, and newly emerging BVLOS operations  

Dr. Sharma is an FAA Part 107 Certified Remote Pilot and serves as an FAA Safety Team (FAAST) representative with a drone specialization, supporting aviation safety outreach and education.  Dr. Sharma is also a certified Tableau Desktop Specialist and Google-certified Data Analyst.

Brooke Wheeler , Florida Institute of Technology

Dr. Wheeler is the Director of the Teaching Assistant Seminar and an Associate Professor in the College of Aeronautics. Her current research covers broad areas of aviation sciences and environmental science, including aviation sustainability, electric aircraft, safety management systems, flight training, and how to grow plants on Mars.

Dr. Wheeler's dissertation focused on how scale of observation, disturbance, and soil nutrients influence plant species diversity. Her research included field study, statistical modeling, and analysis of data in the Carolina Vegetation Survey database. She spent two and a half years as a Postdoctoral Fellow in the Thompson Writing Program at Duke University, teaching and learning about using writing in STEM courses, and helping to run the Writing Studio. Her research interests include writing to learn in STEM courses, flipped classrooms, aviation environmental science, sustainability, environmental communication, and disturbance ecology.

Bhoomin Chauhan , NASA AMES Research Center

I am an experienced, published aviation professional with demonstrated 5+ years of experience in identifying operational and infrastructure considerations, establishing human factors implications, and drafting integration strategies for traditional and emerging technologies – advanced air mobility (AAM), electrical vertical take-off and landing (eVTOL), and unmanned aircraft systems (UAS) into the National Airspace System (NAS).

Shaun Kelly , Purdue University

Shaun Joseph A. Kelly is an aviation executive, educator, and researcher with more than a decade of senior-level airport management experience and a nationally recognized record of leadership in aviation education. He currently serves as an Assistant Professor of Practice in Aviation Management in the School of Aviation and Transportation Technology at Purdue University, where he teaches across undergraduate and graduate programs in aviation operations, safety, security, and management.

Mr. Kelly’s professional background spans complex airport systems, including Part 139 certificated airports, multi-airport authorities, and emergency response operations. His industry experience includes executive and senior leadership roles such as Assistant Airport Director at Brownsville South Padre Island International Airport, Operations Manager at Newport News–Williamsburg International Airport, and Manager of Airport Fire and Operations at Laughlin/Bullhead International Airport. Across these roles, he has led airport operations, safety management systems (SMS), security programs, capital projects, emergency preparedness, federal grant administration, and regulatory compliance initiatives.

As an academic leader, Mr. Kelly has held extensive instructional and administrative roles at Purdue University, Southern Illinois University Carbondale, and Delta State University. His work in higher education includes curriculum development, accreditation leadership, faculty governance, and student mentorship. He has served as an accreditation coordinator for multiple aviation programs and as a Certified Visiting Team Member for the Aviation Accreditation Board International (AABI), participating in accreditation reviews both domestically and internationally.

Mr. Kelly’s scholarship focuses on aviation safety, airport operations efficiency, human factors, and emerging technologies, including artificial intelligence and unmanned aerial systems. He has published peer-reviewed research and is a frequent invited speaker and moderator at national and international aviation conferences.

He is currently a Doctor of Aviation candidate at the Florida Institute of Technology. He holds a Master of Commercial Aviation from Delta State University and a Bachelor of Science in Aviation Business Administration from Embry-Riddle Aeronautical University. His professional credentials include Certified Member (C.M.) of the American Association of Airport Executives, Airport Certified Employee (ACE), Airport Security Coordinator (ASC), International Aerodrome Certified Employee (IACE), and Nationally Registered Emergency Medical Technician (NREMT).

Beyond academia and industry, Mr. Kelly is deeply engaged in professional service and volunteerism, including long-standing leadership roles with the Civil Air Patrol, the FAA Safety Team, and the National Intercollegiate Flying Association. His work reflects a sustained commitment to advancing aviation safety, strengthening the aviation workforce, and bridging the gap between industry practice and academic preparation.