A comparison of classification models in predicting graduate admission decision

Abstract

While the decision process of graduate admissions remains elusive, specific criteria are decidedly significant in determining admission outcome. Prospective students applying to graduate programs experience a real predicament of selecting the right schools to invest limited resources for the application. This paper presents comparisons of various machine learning classification models, including Naïve Bayes, Logistic Regression, Multilayer Perceptron and Decisions Tree models, in predicting the admission outcome of candidates with a set of known parameters using a dataset of 400 applicant records. By comparing the performance metrics of these methods, the study finds Naïve Bayes to be the most accurate model for this type of dataset. Predictive models such as the ones discussed in this paper can be a valuable tool for prospective students in shortlisting universities in their application process. The study also proposes a framework that incorporates machine learning-based classification into the admissions decision process. Implementing such methods may help support graduate admissions committees in streamlining large pools of applications or observing and understanding trends in their past admission decisions.