Significant gender differences in sociocognitive constructs of STEM students: an analysis supported by machine learning techniques (#1093)

Authors

Gutierrez-Rivera, Juan Sebastian

Montoya-Noguera, Silvana

González-Palacio, Liliana

Abstract

This article aims to explore the perceptions of university students enrolled in STEM programs regarding fourteen sociocognitive constructs, with an emphasis on gender differences. The research addresses the persistent gap in female participation in STEM fields, documented both nationally and internationally, which calls for rigorous analytical approaches. The study was conducted at EAFIT University (Medellín, Colombia), based on a sample of 1,007 undergraduate engineering students, from which 665 valid records were analyzed. A structured questionnaire was used, consisting of 97 variables, 67 of which employed Likert-type scales associated with constructs such as self-efficacy, self-regulation, sense of belonging, motivation, among others. The CRISP-DM methodology was applied for the analysis process, including data cleaning and transformation, internal consistency evaluation, inferential statistics, dimensionality reduction through principal component analysis (PCA), and modeling using unsupervised clustering algorithms (K-means and DBSCAN). Evaluation was performed using the silhouette index. The clustering models did not identify well-defined subgroups, showing a concentration of data in a single cluster and low silhouette values (<0.5), suggesting high perceptual homogeneity. However, inferential statistics revealed significant gender differences in 9 out of the 14 constructs. These findings provide a foundation for developing more equitable education policies, based on evidence and focused on the dimensions that reflect actual gaps.