Defining Safe Training Datasets for Machine Learning Models Using Ontologies

<p>Machine Learning (ML) models have been gaining popularity in recent years in a wide variety of domains, including safety-critical domains. While ML models have shown high accuracy in their predictions, they are still considered black boxes, meaning that developers and users do not know how the models make their decisions. While this is simply a nuisance in some domains, in safetycritical domains, this makes ML models difficult to trust. To fully utilize ML models in safetycritical domains, there needs to be a method to improve trust in their safety and accuracy without human experts checking each decision. This research proposes a method to increase trust in ML models used in safety-critical domains by ensuring the safety and completeness of the model’s training dataset. Since most of the complexity of the model is built through training, ensuring the safety of the training dataset could help to increase the trust in the safety of the model. The method proposed in this research uses a domain ontology and an image quality characteristic ontology to validate the domain completeness and image quality robustness of a training dataset. This research also presents an experiment as a proof of concept for this method where ontologies are built for the emergency road vehicle domain.</p>

Name thesis:degree_name

Master of Software Engineering

Level thesis:degree_level

Thesis - Open Access

Discipline thesis:degree_discipline

Electrical Engineering and Computer Science

Year

2023

Author dc:creator

• Vonder Haar, Lynn C

• Machine learning
• safety-critical domain
• trust
• training dataset
• domain ontology
• quality characteristic ontology
• Other Engineering
• Software Engineering

Repository record dc:identifier

https://commons.erau.edu/edt/744

OAI identifier oai:identifier

oai:commons.erau.edu:edt-1766