Higher education institutions continue to face the problem of student attrition, which in turn impacts graduation rates overall. This has numerous drawbacks not only at the university or student levels but has far-reaching influences on society itself (Schuh & Topf, 2012). Although much research has investigated various factors that contribute towards attrition, on average only 40.3% of college students are found to complete their degrees (ACT, 2008). Despite an attempt to better understand the role different kinds of predictors have towards student success (Lotkowski, Robbins, & Noeth, 2004), limited research has assessed to what extent course information adds incremental variability towards predictive modeling of student retention. Lewis and Terry (2016) have investigated the application of multi-level modeling toward such predictors, while data mining techniques have been used sparingly to support the use of differing predictors. For this study, a method of data mining relatively new to the field of educational literature is contrasted with a hierarchically-based statistical approach to support in determining whether any significant course patterns can lead to improved student retention outcomes. Results from the analysis may provide insight into models that contain greater predictive accuracy, with long-term benefits into course placement as more effective advising is applied. Over time, any improved placement is expected to yield positive effects for students and the university as a whole. Keywords: student retention, data mining, symbolic regression, logistic regression, hierarchical analysis, multilevel modeling, statistical techniques, exploratory analysis, area under curve, AUC