MS in Data Science Thesis - Levi
Levi Nicklas successfully completed all the required coursework for his graduate program in data science, and defended his thesis on April 22nd, 2021. The thesis defense was held using the Microsoft Teams platform.
Dr. Sanchez-Arias supervised Levi. Members of the thesis committee Dr. Centeno (DSBA) and Dr. Chintakunta (EE) approved Levi’s thesis work.
Levi’s work explored unsupervised learning methods applied to text data utilizing modern graph based methods as preprocessing. Two datasets, technical reports on ambulance crashes from the NHTSA and social media threads from a reddit forum on mental health, were preprocessed into graphical representations. The graphical representations of the document can then be compared using modern graph kernel methods. The result of applying graph kernels is a matrix of similarity that can be used to perform clustering. The datasets both differ in their writing style and their size.