Predicting Student Success in College Algebra Classes Using Machine Learning

Abstract

College Algebra is a gateway course for STEM majors with large enrollment and low passing rates. We analyze the factors which contribute to student success in College Algebra courses at an urban community college. Characteristics and grades of over twenty thousand students who were enrolled in College Algebra courses between the years 2017 and 2022 have been analyzed. Among the students' characteristics being studied are gender, ethnicity, age, first-generation college status, placement exam scores, grade point averages (GPA), whether they are freshmen or transfer students. The course modalities include online, hybrid or in-person. We study correlations between factors that affect student success. Using k-nearest neighbor and decision tree algorithms, we predict student success based on the student characteristics and course features. Using Chi-Square Test of Independence, we show that passing rates of students depend on gender, ethnicity, age, overall GPA and whether they are freshmen or transfer students. Passing rates also depend on the modality of the course and the semester (fall or spring) the course is taken. With both supervised machine learning algorithms used, the probability of students passing were predicted with approximately 85 percent accuracy. Our results show that machine learning models can successfully be used on student data to predict course outcomes which can enable early intervention to those students with higher chances of failure in the course. Our findings may encourage college administrations to use machine learning for predicting student success and be able to provide better advisement to incoming students regarding course selection.