DescriptionThe goal of this thesis is to demonstrate how machine-learning techniques can be used to improve educational outcomes for STEM students at Rutgers University-Camden. The three main areas of focus are: identifying changes in the academic landscape throughout a 15-year period, identifying predictors of student success, and using these predictors to develop a recommendation system to assist at-risk students. The data in the study con- sists of student demographic and academic records from 2003-2017. Simple exploratory data analysis is used to highlight changes in student performance over time. Next, a deeper analysis is performed by training three classifiers - logistic regression with L1 penalty, logistic regression with L2 penalty, and a random forest model - to predict the probability that students will graduate. Finally, the predictions of each classifier are calibrated and combined to form a robust recommendation system which can be used to alert advisers when a student is struggling.