Hazard Classification of Federal Aviation Administration (FAA) Unmanned Aircraft Systems (UAS) Sightings Reports using Machine Learning
Open AccessThe current rapid progression of Unmanned Aircraft Systems (UASs) introduces a new class of safety risks into the National Airspace System (NAS). Recently, the Federal Aviation Administration (FAA) has witnessed a growing number of UAS sighting reports submitted by pilots, which are written in natural language—unstructured text that requires a human to manually explore the data. Exclusively relying on a human is time-consuming and likely subjective, as humans are inherently biased. This research introduces a new hazard classification scheme and demonstrates an automatic UAS sighting classification by means of Machine Learning (ML) and Natural Language Processing (NLP) techniques. This classification model aims to decrease the time and effort required by aviation analysts to identify and validate UAS safety risks and enhance the efficacy of the risk analysis process. Additionally, this model enables FAA executives to make risk-informed decisions about the potential hazards of UAS integration in the NAS, specifically its impact on the air traffic around airports in the United States. In this research, three ML models and five Deep Learning (DL) architectures were trained to classify 3,132 reports, including Random Forest (RF), Gradient Boosting (GB), Extreme Gradient Boosting (XGB), Artificial Neural Network (ANN), Sequential Long Short-Term Memory (Se-LSTM), Bi-directional Long Short-Term Memory (Bi-LSTM), Convolutional Neural Network (CNN), and Bidirectional Encoder Representations from Transformers (BERT). The results show that BERT scored the highest classification performance with accuracy, precision, recall, and F1-score of 96%, 96%, 96%, and 96%, respectively.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.