Modeling in R and Weka for Course Enrollment Prediction

Abstract

Abstract Predicting course enrollment is a common university resource planning problem. California State University Northridge (CSUN) faces many unique challenges when predicting student enrollment in its undergraduate Computer Science (CS) and Computer Information Technology (CIT) courses. In this paper, we discuss the design of an enrollment prediction tool which ap- plies three Time Series models using R and four Time Series models using Weka utilizing a database of 19 semesters of enrollment data. The seven different models are tested against var- ying amounts of holdout data to see which can best predict enrollment for undergraduate CS and CIT courses to within one standard class size of 25 students. Predictions on holdout data are compared both in modified form, with numbers rounded up and negative values zeroed out, and unmodified form. All models were most accurate when predicting three semesters of hold- out data using the maximum available enrollment data from Spring term of 2010 to Spring term of 2015 for training. The best resulting predictions were accurate within one standard class size of 25 students for 93.5% of Computer Science Department (CSD) courses, and the worst pre- dictions were accurate within one standard class size for 77.4% of CSD courses. Keywords: R, Time Series Forecasting, Weka, Machine Learning.