A baseline admissions prediction model with textual analysis and confidence interval estimations.

Abstract

Essays submitted out individuals applying to Baylor University may contain hidden information that would assist the admissions department in their decision to accept the appicant. Through textual analysis, this paper attempted to reveal signals of a student's intent to attend Baylor if accepted as well as offering an additional platform to judge a student's ability. Both results are independent of other information gathered from the application process. The models created were found not to be strong enough to act as a stand-alone decision rule. However, the new variables created can be used in Baylor's admission model to increase its effectiveness. The groups of words in commitment, Baylor and admissions groups all prove to be influential, which can be used in other parts of the enrollment process, such as phone interviews. This paper also simulates a confidence range around yield estimates generated from the current model being used at Baylor.