Comparing the Performance of Machine Learning Models for Analyzing Categorical Variables in Educational Research

• In Event: Linear, Nonlinear, Latent, and Machine Learning Modeling for Educational Research

Abstract

This study compares commonly used machine learning (ML) techniques for handling
categorical variables. Through synthetic educational data with categorical variables as predictors
for students’ academic performance, we compared the accuracy and model fit of Decision Trees,
Gradient Boosting, CatBoost models, and Multiple Regression. The procedures of the analysis
are demonstrated. This research aims to explore the performance of machine learning models
with synthetic educational data, particularly when dealing with categorical variables as
predictors. By addressing existing challenges and leveraging machine learning techniques, this
investigation seeks to enhance the accuracy and efficiency of estimating outcomes in educational
studies, ultimately empowering educational researchers to make more informed and impactful
decisions.

Authors

• Haiyan Bai, University of Central Florida

• Oluwaseun Peter Farotimi, University of Tennessee

• Fangxing Bai, Montana State University

• Comfort Happiness Omonkhodion, University of Central Florida