Using Machine Learning to Identify Predictors Contributing to the College Enrollment of High School Students With Lower Socioeconomic Status

• In Event: Machine Learning Applications in Education

Abstract

In recent decades, an increasing number of students have enrolled in college, however, students from lower socioeconomic status remain underrepresented. Therefore, the current study attempted to examine actionable factors predict college enrolment among high school students from lower socioeconomic backgrounds using five machine learning algorithms. Participants were 5,223 lower socioeconomic class 9th graders (51% female; mean = 14.59, range = 13~17) from High School Longitudinal Study (HSLS 2009) project. The findings indicated that the random forest was the most accurate classifiers with 70.84% testing accuracy. In addition, the five most powerful predictors of college enrollment were the student's overall GPA, school motivation, school problem, student's sense of school belonging, and student's school engagement.

Authors

• Surina He, University of Alberta
  Presenting Author

• Mehrdad Yousefpoori-Naeim, University of Alberta
  Non-Presenting Author

• Ying Cui, University of Alberta
  Non-Presenting Author

• Maria Cutumisu, McGill University
  Non-Presenting Author