This project expanded the capabilities of NirsAutoML (NAML), an application developed at WPI dedicated to using the sktime library for multivariate time series classification of fNIRS data. NAML was used solely as a command line interface. It was not synced with the sktime library, resulting in the classifiers used in NAML lacking improvements as sktime matured. The application now exposes a remotely hosted frontend for access by researchers, improved code quality for future maintenance, documentation for installation, use, and development, and a Docker container that streamlines the installation and deployment process. A case study for NAML comparing classifier efficacy by channel feature type has been performed on the “fNIRS to Mental Workload” dataset provided by Tufts University [1].

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Houlihan, Patrick

Publisher

• Worcester Polytechnic Institute

Identifier

• 64816
• E-project-042722-231506

Keyword

• Machine Learning
• Container
• fNIRS
• Brain Data
• Human Computer Interaction

Advisor

• Solovey, Erin
• Neamtu, Rodica

Year

• 2022

Date created

• 2022-04-27

Resource type

• Major Qualifying Project

Major

• Computer Science

Rights statement

• In Copyright