Enrollment Prediction through Data Mining

In this paper, we describe our study on enrollment prediction using support vector machines and rule-based predictive models. The goal is to predict the total enrollment headcount that is composed of new (freshman and transfer), continued and returned students. The proposed approach builds predictive models for new, continued and returned students, respectively first, and then aggregates their predictive results from which the model for the total headcount is generated. The types of data utilized during the mining process include population, employment, tuition and fees, household income, high school graduates, and historical enrollment data. Support vector machines produce the initial predictive results, which are then used by a tool called Cubist to generate easy-to-understand rule-based predictive models. Finally we present some empirical results on enrollment prediction for computer science students at California State University, Sacramento