Analyzing College Students' Advising Records To Improve Retention And Graduation Outcome

• Main Authors: Yang, Hui, Pimparkar, Apurva D., Graterol, Celia, Kased, Rama Ali, Love, Mary Beth
• Format: Conference Proceeding
• Language: English
• Subjects: Academic advising; Conferences; Data mining; Educational data mining; Engineering profession; Graduation rate; Machine learning; Natural language processing; Retention rate; Text mining; Timing
• ISSN: 2377-634X
• DOI: 10.1109/FIE49875.2021.9637129

This full research paper studies the impact of faculty-student advising interactions over students' retention and graduation outcomes. These 3000+ students were enrolled in a community-based learning program between 2014 and 2018 at San Francisco State University, a 4-year public degree-granting university. To address the ongoing challenge of low retention and graduation rates, this program has introduced early and preventative interventions through both mandatory and voluntary advising with a team of 40+ advisors. We apply statistical and machine learning approaches to analyze the advising records generated over a period of 5 years. Specifically, we have addressed the following problems: (1) what is the impact of timing and frequency of advising interactions on students' retention and graduation status? (2) how effective are such interactions in helping academically at-risk students? (3) What are the common types of interactions and how are they correlated with students' academic performance? Finally, (4) how much insight can we further gain through looking into advisors' written notes? Our results show that the current advising practice is not biased against a specific demographic group. They also show that it is important to require advising during a student's first term of her college career. Delaying such interactions to the second term, for instance, can significantly lower the likelihood to retain students in their third term and then seventh term, consequently reducing their chance to graduate within four or five years. Finally, after having applied a host of natural language processing (NLP) techniques (e.g., LDA topic modeling) to advisors' written notes, the results suggest that a more rigorous note-taking approach is needed in order for such notes to help gain broader and deeper insights into students' advising needs. Our results show that the current advising practice is not biased against a specific demographic group. They also show that it is important to require advising during a student's first term of her college career. Delaying such interactions to the second term, for instance, can significantly lower the likelihood to retain students in their third term and then seventh term, consequently reducing their chance to graduate within four or five years. Finally, after having applied a host of natural language processing (NLP) techniques (e.g., LDA topic modeling) to advisors' written notes, the results suggest that a more rigorous note-tak