Project AVIAN: Implications of Utilizing the Novel AVIAN-S Machine Learning Model in Analyzing Aviation Safety Event Reports

Abstract

Voluntary Safety Reporting Programs (VSRPs) create an opportunity for actively identifying potential safety issues within aviation operations. However, manually reading and analyzing these reports can be labor-intensive and heavily relies on subject-matter experts. The full potential of VSRP data is difficult to achieve due to limited resources available to extract meaningful human factors (HF) data from reports. New machine learning (ML) techniques involving natural language processing (NLP) offer opportunities to label factors of interest within safety reports more efficiently and effectively. A novel ML model was developed to identify HF issues within aviation safety reports. The AVIAN-S model was trained on over 50,000 rows of manually classified VSRPs. The model uses ML and NLP to automate the process of labeling aviation safety reporting data and coding reporter narratives according to an established HF taxonomy. Preliminary results and implications for using the AVIAN-S model for analyzing aviation safety event reports are described.