Computer Science (CS) is a competitive field with high demand and low admission rates. In the last three years, the CS program at the University of California, Irvine (UCI) received over 15,000 applications and admitted only 14%. Despite being highly selective and admitting top students, we observed over a 10-year span 57% experienced a period of academic probation. The alarmingly high probation rate motivates us to understand challenges that lead CS students to enter probation at UCI. Some of our results aligned with past findings regarding academic performance: high school GPA and math background level have a negative trend with probation, underrepresented groups experience probation at higher rates than the average, and female students have lower probation but higher attrition rates than men. More importantly, over a third of the students enter probation in freshman year, and students on probation leave the field at higher rates than the average. There is also a positive trend linking probation duration to attrition rates. Our results suggest that current probation practices may not be sufficient for students to return to satisfactory academic standing, highlighting the need for proactive and targeted probation interventions.

Application processes in higher education, particu-larly in fields like Computer Science (CS), have become more competitive in recent years. Additionally, at the University of California, Irvine (U CI), 57 % of students who were admitted and enrolled in the CS program between Fall 2011 and Fall 2015 failed to maintain satisfactory academic standing and entered academic probation. Of these probation students, 41 % entered probation during their first year. The objective of this study is to explore the feasibility of using machine learning tools to identify the academic probation status of first-year CS students at an early stage, with the ultimate aim of providing timely support to those who are at the greatest risk of entering academic probation. We trained various classifiers on a dataset containing demo-graphic and academic performance variables for CS students at U CI, and performed feature analysis to identify the most influential variables. Since it is our priority to identify students who enter academic probation, our focus was on maximizing recall. The best-performing classifiers were Linear Support Vector Machines (recall of 0.838) and Logistic Regression (recall of 0.812). Our feature analysis revealed that ethnicity, student region, and having taken the Advanced Placement CS exam had a greater impact on the classification task than GPA and summer course enrollment. Future work could focus on expanding the dataset to include academic data available after enrollment to improve our current results. Our findings could also be applied to assess their impact in strategies for student retention in CS. Overall, this study demonstrates the potential for machine learning tools to assist in identifying at-risk students in CS, and highlights the importance of considering a range of demographic and academic variables in this process.

Pursuing higher education is a competitive process. Many students dropping out risk having less career opportunities. We implemented machine learning (ML) models to identify students at risk of probation and discover new factors correlated with probation cases. This would allow to proactively provide students at risk with support to maximize academic success, and also propose curricula changes.