Machine Learning Models of Participation in Work-Related Training
Open Access
- Author:
- Zigner, Michael
- Graduate Program:
- Workforce Education and Development
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 15, 2021
- Committee Members:
- David Lynn Passmore, Dissertation Advisor/Co-Advisor
David Lynn Passmore, Committee Chair/Co-Chair
Michael W Harvey, Committee Member
Cynthia Pellock, Committee Member
Jonna Marie Kulikowich, Outside Member
Heather Toomey Zimmerman, Program Head/Chair - Keywords:
- workforce development
human resource development
human resources analytics
machine learning
predictive modeling
work related training
training and development - Abstract:
- Predictive modeling has become so ubiquitous that it now impacts our daily lives. These models can be seen at work in determining the ads we see on social media, controlling our e-mail spam, and preparing us for extreme weather events. The predictions from these types of analyses have been shown to be powerful and valuable, and are rapidly being adopted for new and innovative purposes as the technology becomes available to implement them. Many academic disciplines now use predictive models as part of their regular research practices; and the number of academic articles that use this methodology continue to increase yearly. However, the adoption of this innovation has yet to happen in workforce development-related fields. This is unfortunate because by not adopting an analytical mindset and building technological skills, practitioners face being left behind and marginalized within their organizations. The choice to understand and utilize innovations like predictive modeling is rapidly becoming a necessity rather than a luxury for professionals working in the field. An underlying purpose to this study, therefore, is to foster the continued adoption of advanced analytics like predictive modeling among workforce development researchers and practitioners. By demonstrating the building and evaluation of predictive models using a topic germane to the field (i.e., work-related training), this study helps provide the how-to and principles knowledge needed to spur adoption (Rogers, 1995), and provide confirmation of its value by subject-matter experts. This study used a large, publicly available data set (the National Survey of College Graduates; n = 275,598) and more than 150 variables to build and evaluate a selection of predictive models that can accurately identify future participants in work-related training. This document walks the reader through the predictive modeling process. It explains key concepts and principles that help build a foundational understanding of the strengths and weaknesses of the different models that are necessary for an effective evaluation of a model’s predictions. Most importantly, by providing a detailed demonstration of the predictive modeling process on a relevant topic, this dissertation provides a familiar context that helps the practitioner understand how this new innovation applies to their profession.