Is this project an undergraduate, graduate, or faculty project?

Graduate

Project Type

group

Campus

Daytona Beach

Authors' Class Standing

Isaac Morrison, Graduate student Kaitlyn Cavanaugh, Graduate student

Lead Presenter's Name

Isaac Morrison

Lead Presenter's College

DB College of Arts and Sciences

Faculty Mentor Name

Dr. Gamage Dumindu Samaraweera

Abstract

Aviation accidents are rarely the result of a single failure but rather from a complex causal chain of latent failures. While traditional data mining models often predict incident occurrence, they frequently overlook the sequential mechanics defined by known accident causation theoretical frameworks like the Swiss Cheese Model and the FAA's HFACS. This project addresses the need for interpretable, reliable, multi-stage forecasting by proposing a Sequential Causal Architecture that transforms theoretical causation models into a structured Directed Acyclic Graph (DAG) for multimodal accident causation chain prediction. Data from the NTSB and DOT is used and connected together in a meaningful way to integrate diverse attributes that align with the designed accident causation DAG. A comparative analysis of three different data mining approaches: Random Forest, Bayesian Networks, and Long Short-Term Memory (LSTM), is conducted, maintaining causal integrity in each implementation. Models are holistically evaluated to enable trust within the aerospace safety community with evaluations such as accuracy, weighted F1-score, feature importance, and sensitivity. By shifting focus on predicting "what" happened to "how" and "why" it happened, this project intends to provide an explainable and reliable machine learning model for proactive and actionable risk mitigation in aerospace safety management systems.

Did this research project receive funding support (Spark, SURF, Research Abroad, Student Internal Grants, Collaborative, Climbing, or Ignite Grants) from the Office of Undergraduate Research?

No

Share

COinS
 

Sequential Causal Architecture for Multimodal Aviation Accident Prediction

Aviation accidents are rarely the result of a single failure but rather from a complex causal chain of latent failures. While traditional data mining models often predict incident occurrence, they frequently overlook the sequential mechanics defined by known accident causation theoretical frameworks like the Swiss Cheese Model and the FAA's HFACS. This project addresses the need for interpretable, reliable, multi-stage forecasting by proposing a Sequential Causal Architecture that transforms theoretical causation models into a structured Directed Acyclic Graph (DAG) for multimodal accident causation chain prediction. Data from the NTSB and DOT is used and connected together in a meaningful way to integrate diverse attributes that align with the designed accident causation DAG. A comparative analysis of three different data mining approaches: Random Forest, Bayesian Networks, and Long Short-Term Memory (LSTM), is conducted, maintaining causal integrity in each implementation. Models are holistically evaluated to enable trust within the aerospace safety community with evaluations such as accuracy, weighted F1-score, feature importance, and sensitivity. By shifting focus on predicting "what" happened to "how" and "why" it happened, this project intends to provide an explainable and reliable machine learning model for proactive and actionable risk mitigation in aerospace safety management systems.

 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.