Multivariate Analysis of the Performance of Industrial Engineering Programs in Colombia Based on Saber Pro Test

Abstract

This study compares the academic performance profiles of 84 Industrial Engineering academic programs in Colombia based on results from the 2024 national standardized higher education test (Saber Pro). The main objective is to progress from an average to a classification process for academic programs, incorporating heterogeneous practices across academic competencies. The methodological pipeline is constructed on unsupervised machine learning methods: UMAP (Uniform Manifold Approximation and Projection) was used to reduce the dimensionality of the data by keeping the complex local and global structure and K-means classification was used to group the academic programs. The optimal final number of clusters was calculated to be k  = 3, determined both by elbow analysis and the average silhouette index. The programs were successfully clustered into three groups: high-performing (Cluster 1), intermediate-performing (Cluster 3), and low-performing (Cluster 2). MANOVA and univariate ANOVAs were subsequently used to characterize the magnitude of between-cluster differences across the eight competencies ( p  < 0.001); because the clusters were derived from these same variables, these analyses describe the resulting groups rather than providing an independent validation of the clustering solution. The analysis showed that the largest cluster differences were observed in English Language and Quantitative Reasoning. In contrast, Written Communication and Scientific Thinking – Mathematics and Statistics showed smaller relative differences. This analysis is intended to provide an exploratory multivariate method for profiling academic performance in higher education. The resulting classification represents a potential input for academic program management and for public policy discussions on equity and academic achievement in Colombian Industrial Engineering education; such applications, however, would require prior validation of the clusters against external criteria.