← All work
03 · 2025
Student Lifestyle & Academic Performance
Behavioral analysis and a multinomial classifier predicting GPA tiers from study habits
Role
Solo analyst + modeler
Year
2025
Stack
PythonPandasNumPyScikit-learnSeabornPower BI
01
Overview
How much do everyday behaviors — study time, sleep, stress, social hours — actually predict academic outcomes?
The goal was twofold: a defensible classifier of GPA tiers, and a clear behavioral story a non-technical reader could follow.
03
Results
64%
overall accuracy across GPA tiers
0.74
F1-score (High GPA tier)
Study hours
strongest performance predictor
09
Reflection
GPA went up with study time and down with stress — predictable, but the magnitude mattered. The fairness check was the part that changed how I think about modeling: 'minimal disparity' is a result worth writing down, not an afterthought.”