Predicting school transition rates in Austria with classification trees

TL;DR

This paper demonstrates that classification trees can effectively analyze educational data in Austria, selecting relevant variables, matching traditional models in performance, and offering a data-driven approach in educational sciences.

Contribution

It introduces the application of classification trees to educational data, showing their ability to select variables aligned with theories and perform comparably to regression models.

Findings

Trees select variables consistent with educational theories.

Classification performance is comparable to regression models.

Trees assist in variable selection for regression analysis.

Abstract

Methods based on machine learning become increasingly popular in many areas as they allow models to be fitted in a highly-data driven fashion, and often show comparable or even increased performance in comparison to classical methods. However, in the area of educational sciences the application of machine learning is still quite uncommon. This work investigates the benefit of using classification trees for analyzing data from educational sciences. An application to data on school transition rates in Austria indicates different aspects of interest in the context of educational sciences: (i) the trees select variables for predicting school transition rates in a data-driven fashion which are well in accordance with existing confirmatory theories from educational sciences, (ii) trees can be employed for performing variable selection for regression models, (iii) the classification performance of trees is comparable to that of binary regression models. These results indicate that trees and possibly other machine learning methods may also be helpful to explore high-dimensional educational data sets, especially where no confirmatory theories have been developed yet.