Cognitive and Non-Cognitive Predictors of Educational Profile Choice: An Analysis of Fields of Study Using Machine Learning

Authors:
Open Access
Publisher website
Google Scholar
Abstract
This article presents the results of a study on cognitive and non-cognitive predictors of university educational profile choice. The sample included 6,290 students from 23 Russian univer- sities (37.6% males, 62.4% females, aged 17–24 years). Academic fields were aggregated into five macro-groups: humanities, natural sciences, pedagogy, social sciences and economics, and technical sciences. Cognitive abilities were assessed using intelligence tests (Raven’s matrices), working mem- ory, spatial abilities (rotation, visualization, relations), and reaction time. Non-cognitive character- istics were measured using questionnaires for math anxiety, mental toughness, and psychological well-being. Statistical analysis included multifactorial ANCOVA and predictive modeling using ran- dom forest. The results showed that the field of study is associated with differences in cognitive and personality profiles, with spatial abilities contributing the most (η 2 up to 0.030). Students in techni- cal fields demonstrated higher cognitive performance, while humanities students exhibited lower well-being and higher math anxiety. Gender significantly affected spatial abilities and non-cognitive characteristics but not general intelligence or working memory. The combined machine learning model achieved an accuracy of 39.7% (against a baseline of 29.1%). A discrepancy between actual and desired major choice was found to be primarily related to cognitive mismatch. Gender-specific patterns were identified: spatial abilities play a leading role for males, whereas general intelligence and psychological well-being are more important for females. These findings can be applied in career guidance and educational practices.

This publication has 31 references indexed in Scilit: