Movement Matters: Investigating The Correlation Between Student Behavior and Academic Performance – A Pilot Study

This study examines the relationship between observable behavioral patterns and academic performance among first-year BS Data Science students during Python programming exercises. From an initial dataset of 24 recorded videos of students completing the same programming task, five videos were selected for analysis. Using a combination of MediaPipe-based computer vision analysis and human annotation, behavioral indicators-including facial expressions, attention patterns, and body posture-were systematically analyzed. A structured annotation methodology was implemented, achieving high inter-rater reliability among human annotators (mean$\kappa=0.763-0.825$) and moderate to substantial agreement between MediaPipe and human observations (mean$\kappa=0.544-0.665$). Results revealed significant correlations between behavioral patterns and academic performance, with attention ratio ($\mathrm{r}=0.896, \mathrm{p}