The goal of this research and development project is to further explore the potential of Machine Learning (ML) to improve learning outcomes through the use of xAPI data by early prediction of at-risk students. This research has the potential to significantly impact the learning community by promoting early intervention, personalized learning approaches, efficient resource allocation, continuous improvement, and enhanced teaching practices, ultimately leading to better outcomes for all students. Early results from Phase 1 of this research reported at I/ITSEC 2023 yielded a very promising 78% accuracy. Phase 2 of this research will continue investigating various windows of sequences between pass/fail events based on a larger xAPI dataset of approximately 200 Million data points (xAPI statements) from a DoD component. Overlapping windows of shorter sequences from between events will be tested to see if predictions can be successful ignoring later data in a sequence. While we have developed a promising prediction model for student outcomes, we have not yet discovered the root causes for student failures based solely on sequences. Further research will extract the specific attributes of a sequence that drives the prediction decision. The discovery of such features can aid in intervention decisions. One intervention recommendation may be for a student to limit the number of simultaneous courses. Another intervention recommendation may be for a student to complete more course exercises before taking assessments. There are several techniques for discovering features such as visualization observations, permutation feature importance, Shapley Additive Explanations, Attention Mechanism, and logistical regression. This research aims to provide instructors with the tools to manage large populations of students that would be difficult to guide through a learning curriculum without the aid of ML techniques and resulting knowledge.