Cybersecurity researchers have made significant steps to understand the mechanisms of security policy compliance and unify theories of security behavior. However, due partly to the limitations of traditional variance model statistical methods, these studies by necessity typically focus on a single security policy issue. By contrast, new machine learning algorithms frequently employed by data scientists offer great promise as a new statistical approach for examining robust individualized interpretations of policy and can also identify potentially risky behaviors. This study proposes to explore cybersecurity training impediments of multiple protection motivation behaviors in ransomware prevention training. It demonstrates the feasibility of using machine learning with survey items from the cybersecurity research to predict non-compliance. It also illustrates a potentially novel method to statistically validate research theory through higher levels of ML prediction. This study is a work in progress and we seek feedback on its design and relevance.