Creating a Model to Predict Student Success using WeBWorK data

Student success is a major focus in the educational system, where a variety of predictors are used to estimate and measure how well students do in their different classes at the end of the academic year. Our research project aims towards proposing a model capable of demonstrating how student success can be predicted based on a series of indicators gathered from work submitted by the student throughout the semester. We studied the student’s performance in the open-source online homework assignment system WeBWorK for a mathematics course, taking into account the final score in a given assignment, and the number of times every problem was tried by the student before obtaining a correct answer. Data from one Pre-Calculus and two Calculus I courses at St Thomas University was used to create a multinomial logistic regression model that takes into account the student’s scores in all assignments during a semester, as well as the student’s “success index” per assignment, a fairly good indicator of how well the student is grasping the concepts evaluated in every assignment.