Predicting time to graduation at a large enrollment American university

Abstract

The time it takes a student to graduate with a university degree is mitigated by a variety of factors such as their background, the academic performance at university, and their integration into the social communities of the university they attend. Different universities have different populations, student services, instruction styles, and degree programs, however, they all collect institutional data. This study presents data for 160,933 students attending a large American research university. The data includes performance, enrollment, demographics, and preparation features. Discrete time hazard models for the time-to-graduation are presented in the context of Tinto’s Theory of Drop Out. Additionally, a novel machine learning method: gradient boosted trees, is applied and compared to the typical maximum likelihood method. We demonstrate that enrollment factors (such as changing a major) lead to greater increases in model predictive performance of when a student graduates than performance factors (such as grades) or preparation (such as high school GPA).

Targeted projection pursuit similarity based attribute selection for academic performance prediction

“…Based on model learning management system, ontology method was introduced 19 to develop the transportable predictive models for students' performance. The logistic regression gradient boosting was introduced ( 20 ) to provide better fits than traditional maximum likelihood.…”

A Framework for Evaluating Statistical Models in Physics Education Research

“…Statistical modeling allows for a process of rigorous testing of these theories. In reverse, theory allows for strong explanations about when a model fails (e.g., a lack of social network data can impact the overall predictability when examining university student drop out [40]).…”

“…Additionally, more sophisticated models can increase our understanding even when we have lower amounts of data on student sub-populations. For example, using a data set that had a small racial and ethnic minority population ( 15%) Aiken et al [40] presented gradient boosted logistic regression models that performed much better than traditionally solved maximum likelihood models when predicting undergraduate student time to graduation. Statistical model choice can impact not only the overall predictability of a model, but also our understanding of a system even when our data on all populations are not equal.…”

“…Models may also not be able to be generalized beyond a local context. Newer methods, such as boosting, also have the ability to increase understanding of sub-populations that traditionally are poorly fit by statistical models due to lack of data [40]. Ultimately, there is not one "true" model, but many different models that can inform us on the dynamics of students in educational systems in physics.…”

Predicting On-time Graduation based on Student Performance in Core Introductory Computing Courses using Decision Tree Algorithm

“…A large number of students enroll in college each year, yet many of them fail or leave out in less than three years (Orion, Forosuelo & Cavalida, 2014). Understanding the paths students take to complete their degrees can assist instructors and administrators better serve student populations and help them achieve their educational objectives (Aiken et al, 2020). Universities must be concerned with the creation of effective mechanisms for monitoring students' progress and identifying crucial components of their performance in order to reach a higher level of education quality (Tampakas et al, 2018).…”