Predicting Student Retention Using Support Vector Machines

Universities and colleges have a constant focus on improving student retention and degree completion rates. Degree completion impacts the reputation of the institution, as it is a reflection of institutional ability to prepare graduates with the specific skills that contribute to society through STEM majors. Colleges and universities collect considerable amounts of student data; however, efforts need to be made to utilize the data to increase student success. For instance, by determining the factors that influence student retention and completion rates, it is possible to improve advising through intentional student advising. To this end, this research presents the application of support vector machines (SVM) to predict degree completion within three years by STEM community college students. SVM enables the classification of the input variables into expected classes, completion and not completion, by maximizing the margin between the points from the different classes constraining the misclassification. The model was developed using data on 282 students with 9 variables. The variables included age, gender, degree, and college GPA, among others. The model results, which include prediction and variables ranking, offer an important understanding about how to develop a more efficient and responsive system to support students.