Leveraging Educational Analytics to Identify Cognitive and Behavioral Influences on Adolescent Learning

Abstract

The incorporation of Cognitive Behavioral Therapy (CBT) philosophies with machine learning (ML) signifies a very promising interdisciplinary method to analyse the adolescent academic performance. This research study studies whether the CBT-derived psychological constructs the resilience, emotional regulation, anxiety management, cognitive distortions, and as well as metacognitive awareness. This can meaningfully contribute to predict the academic conclusions if we combine with traditional educational indicators. A dataset of 300 secondary school students from ages 18–22, mean age 20.6 years) was analysed utilising a newly technologically advanced 40-item CBT-based Adolescent Assessment Scale (CBT-AAS-40) alongside attendance, study hours, as well as academic percentage scores. The CBT-AAS-40 have proved acceptable internal consistency (Cronbach's α = 0.82 for total scale, α = 0.74–0.79 for subscales). Total five regression algorithms were compared including Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Regression (SVR). Model performance was evaluated using R², MAE, and RMSE on an 80/20 train-test split with 5-fold cross-validation. Random Forest achieved the best performance (R² = 0.58, MAE = 5.23, RMSE = 6.87), followed by Gradient Boosting (R² = 0.55) and SVR (R² = 0.51), while Linear Regression yielded R² = 0.42. Attendance emerged as the strongest predictor (r = 0.52, p < 0.001), followed by resilience score (r = 0.41, p < 0.001) and study hours (r = 0.38, p < 0.001). Resilience-related features were the most influential psychological predictors, supporting the "Resilience-Engagement" hypothesis. These findings determine that CBT-informed constructs contribute meaningful explanatory power beyond traditional educational indicators, to explain approximately 58% of academic performance variance when combined with attendance and study behaviour. The research study contributes to the emerging field of computational psychoeducation by showing how ML can identify specific cognitive levers within CBT frameworks that associate with real-world educational outcomes.