Abstract:
State and federal governments regularly focus on improving student retention and completion in higher education as a means of increasing the skills of the workforce to better meet the challenges of a global economy. The findings of this research present a statewide picture of retention for nontraditional students in the Technical College System of Georgia and generalizations could be used to specifically improve processes and procedures on how colleges recruit and respond to this growing and diverse student population. With a specific focus on nontraditional students in diploma and certificate programs, the outcomes of this research will allow decision-makers to consider how student factors, and the relationship between those factors, influence nontraditional student progression in order to make informed decisions on how to better serve the needs of this specific student population.
The purpose of this nonexperimental, ex post facto, correlational study was to examine the predictability of academic factors (student GPA and program type), background factors (age, race or ethnicity, gender, high school diploma type, high school graduation date), and environmental factors (Pell eligibility, single parent status, displaced homemaker status) on the retention of nontraditional students enrolled in diploma and certificate programs in the Technical College System of Georgia. To do so, this study addressed which prediction model, out of two data modeling approaches (logistic regression and linear discriminant analysis) and three data mining approaches (classification tree, random forest, and support vector machine models), best predicts whether a student was retained or not retained.
The predictor variables GPA, programs related to Transportation and Logistics, female students, Black students, and Pell eligibility were influential in students being retained. Being out of high school for five years or more and being enrolled in Cyber, Engineer, or Healthcare programs or Industrial Technology programs were influential predictors of students not being retained. The support vector machine will generate an accurate classification model based on the goal of correctly identifying students who will not be retained so adequate assistance and resources can be provided to them.
Keywords: Retention, Nontraditional, Technical College System of Georgia, Prediction Model, Diploma Programs, Certificate Programs