PCA-Based Dimensionality Reduction and Logistic Regression Modeling of Students’ GPA

Na’imah Hijriati (Universitas Lambung Mangkurat, Indonesia)
Rikha Syahda (Lambung Mangkurat University, Indonesia)
Aprida Siska Lestia (Lambung Mangkurat University, Indonesia)
Mochammad Idris (Lambung Mangkurat University, Indonesia)
Muhammad Ali Rizqan (Universitas Lambung Mangkurat, Indonesia)
Abdul Rasyid Nasyar (Universitas Lambung Mangkurat, Indonesia)

FIBONACCI: Jurnal Pendidikan Matematika dan Matematika
Vol. 12 No. 1 (2026): FIBONACCI: Jurnal Pendidikan Matematika dan Matematika

This study investigated the effects of socioeconomic, motivational, and psychological factors on students’ academic performance using dimensionality reduction and predictive modeling. Principal Component Analysis was applied to reduce 29 observed variables into latent components based on eigenvalues and explained variance criteria. Logistic regression was then used to model the probability of achieving a high Grade Point Average (GPA). The results showed that psychological components were the most significant predictors, with four out of five psychological components being statistically significant, while only one socioeconomic component was significant. The model demonstrated good fit with a Nagelkerke  of  and classification accuracy of . These findings indicated that psychological support played a dominant role in predicting academic performance, suggesting that interventions focusing on mental and emotional factors could improve students’ academic outcomes.