A Machine-Learning-Based Approach to Informing Student Admission Decisions.

Area of Science:

• Higher Education Management
• Data Science in Education
• Predictive Analytics

Background:

• University admissions face challenges with high application volumes and limited study places, necessitating strategic management.
• Traditional methods using historical enrollment yields ignore statistical uncertainty, leading to suboptimal admission decisions and potential over- or underenrollment.

Purpose of the Study:

• To develop and evaluate a novel machine learning-based approach for optimizing student admission decisions.
• To improve the accuracy of enrollment predictions by incorporating statistical uncertainty.

Main Methods:

• Trained and compared multiple machine learning models on historical university application data.
• Developed a model to predict enrolled applicants conditionally, considering statistical uncertainty.
• Applied the best model to estimate individual enrollment probabilities and aggregated these to predict total enrollment and associated risks.

Main Results:

• The proposed machine learning approach demonstrated superior performance over traditional methods.
• Enabled data-driven adjustments to the number of admitted applicants.
• Effectively controlled the risk of over- and underenrollment.

Conclusions:

• The machine learning-based approach offers a more robust and data-driven solution for strategic student admission management.
• This method enhances the precision of enrollment predictions, leading to more efficient resource allocation and improved student intake outcomes.