Exploring heterogeneous profiles of academic performance among college students: a latent profile and LASSO regression analysis.

Objective: With the development of higher education, college students face various psychological pressures and challenges. Consequently, academic performance has become a key focus of educational research. This study aims to identify latent categories of academic performance among college students and further explore risk factors associated with these categories. Methods: Employing a convenience sampling strategy, 548 undergraduate students were recruited from universities in Henan Province between May and July 2025, the effective response rate is 94%. Data collection utilized the Family Affluence Scale-III, General Self-Efficacy Scale, Perceived Social Support Scale, Academic Performance Scale, General Procrastination Scale, Utrecht Work Engagement Scale-Student, and demographic questionnaire. Three academic performance-related scales served as observed variables. Latent profile analysis (LPA) identified latent categories of academic performance among college students. Univariate analyses using Chi-square tests, Mann-Whitney U tests, or Fisher's exact tests compared characteristics and differences across academic performance categories. Collected variables were incorporated into a relaxed least absolute shrinkage and selection operator (LASSO) method based on bootstrap resampling to confirm key factors influencing these categories. Logistic regression analysis was then conducted on the identified factors. Results: Two latent categories of academic performance among college students were identified: "Low Self-Regulators" and "High Self-Regulators." After variable selection via Bootstrap Relaxed LASSO regression, with the "High Self-Regulators" serving as the reference group, logistic regression analysis revealed that general self-efficacy (OR = 0.79, 95% CI = [0.73-0.85], Conclusion: This study adopted a person-centered approach and identified two latent categories of college students' academic performance through latent profile analysis: "Low Self-regulators" (average academic performance and engagement, higher procrastination) and "High Self-regulators" (higher academic performance and engagement, lower procrastination). Further screening via Bootstrap relaxed LASSO regression revealed that general self-efficacy, perceived social support, and purposeful exercise frequency were the most stable predictors of student category membership. These findings provide empirical support for implementing differentiated academic interventions in higher education, suggesting educators should precisely enhance students' academic adaptation and effectiveness by addressing the psychological needs of distinct groups through boosting self-efficacy and strengthening social support systems.