Understanding Student Academic Performance

An interactive machine learning visualization for CS-GY 9223: Visualization for Machine Learning

999

Students

27

Features

8

ML Models

0.82

Best R² Score

Exploratory Data Analysis

Dataset Overview

Total Records

999

Total Features

27

Target Variable

Semester GPA

Feature Categories

Academic

  • Previous Semester GPA
  • Attendance Percentage
  • English Proficiency Score

Behavioral

  • Study Hours/Week
  • Sleep Hours/Day
  • Screen Time
  • Library Usage

Wellbeing

  • Stress Level
  • Homesickness
  • Social Support

Demographic

  • Gender, Race/Ethnicity
  • Country of Origin
  • Visa Type

Feature Correlation with GPA

Which factors are most strongly associated with semester GPA?

Student Embedding (UMAP / t-SNE)

Visualize student clusters — brush to select students and link to other charts.

GPA Distribution

🟢 Positive Factors

🔴 Negative Factors

Machine Learning Models

Model Performance Comparison

We trained 8 different regression models to predict semester GPA. Here's how they compare:

🏆 Best Model

Not available

R² Score (Test) Not available
RMSE (Test) Not available
MAE Not available

Model performance data not available.

Not available

R² Score Not available
RMSE Not available

Not available

R² Score Not available
RMSE Not available

R² Scores Across Models

RMSE Across Models

SHAP Summary (Model Explainability)

If available, SHAP mean absolute values show per-feature contribution to predictions.

Feature Importance (Top 10)

Which features does our best model rely on most for predictions?

Model Insights

Previous GPA is Key

Previous semester GPA is the strongest predictor of current performance, showing strong academic continuity.

Study Hours Matter

Students who study 25+ hours per week show significantly higher GPAs than those studying <10 hours.

Sleep is Critical

Sleep hours are inversely correlated with stress and strongly related to academic performance.

Key Findings & Patterns

📊 The GPA Distribution

Mean GPA 2.52
Median GPA 2.51
Std Dev 0.43
Range 1.98 - 3.98

🎯 Key Success Factors

  1. Previous GPA (r=0.89) - Historical performance is the best predictor
  2. Study Hours (r=0.67) - More study time = higher GPA
  3. Attendance (r=0.58) - Showing up matters significantly
  4. Sleep Hours (r=0.45) - Well-rested students perform better
  5. English Proficiency (r=0.42) - Language skills support success

⚠️ Risk Indicators

High Stress

Students with stress levels >8 have 0.31 lower GPA on average

Low Attendance

Attendance <60% correlates with 0.48 lower GPA

Inadequate Sleep

Students sleeping <5 hours show higher stress and lower performance

💡 Interesting Patterns

Geographic Diversity

Students from different countries show varying patterns in work hours, family support, and stress levels, but GPA variations are explained more by behavioral factors.

Financial Impact

Higher family income correlates with better access to resources, but scholarship students often have high motivation and competitive GPAs.

Work-Life Balance

Students working 15+ hours/week show lower study hours, but some manage to maintain high GPA through efficient time management.

Interactive GPA Predictor

Adjust student characteristics to see predicted semester GPA

2.5
20 hours
75%
6 hours
5
85

Adjust the sliders and click "Calculate Predicted GPA" to see results

Sample Student Profiles

Course Application: CS-GY 9223

Visualization for Machine Learning

Visualization Techniques Applied

  • Correlation Heatmaps - Understanding feature relationships
  • Distribution Charts - Analyzing variable patterns
  • Model Comparison - Visual performance evaluation
  • Feature Importance - Identifying key predictors
  • Interactive Dashboards - Real-time prediction exploration

ML Model Insights

  • Regression Models - Continuous target prediction
  • Model Assessment - R² score, RMSE evaluation
  • Feature Analysis - Importance and correlation
  • Model Interpretability - Understanding predictions
  • Visual Analytics - Actionable insights