A comprehensive end-to-end analysis of student academic performance using the UCI Student Performance dataset.
Covers preprocessing → correlation → lifestyle → statistical hypothesis testing → interactive Streamlit dashboard.

📊 View Dashboard · 📁 Explore Modules · 💡 Showcase Insights · 🚀 Quick Start

---

📌 Project Overview

This project investigates what factors most significantly influence student academic performance — specifically final grades (G3) in Mathematics — using real-world data from Portuguese secondary schools.

The analysis spans 7 Python modules, each targeting a specific dimension of student life:

Dimension	Question Asked
📖 Study Time	Does studying more hours per week lead to better grades?
👨‍👩‍🎓 Parental Education	Do students with more educated parents perform better?
⚥ Gender	Is there a statistically significant grade gap between genders?
🔗 Correlation	Which variables are the strongest predictors of G3?
🍺 Lifestyle	How do alcohol consumption and social habits affect performance?
🧪 Statistics	Can ANOVA and t-tests validate these observations?
💡 Policy	What interventions can schools implement based on this data?

---

🗂 Dataset Description

Property	Detail
Source	UCI Machine Learning Repository
Original Files	student-mat.csv (Math) · student-por.csv (Portuguese)
Working File	student-mat-clean.csv (cleaned & preprocessed)
Records	395 students
Features	33 attributes (demographic, social, academic)
Target Variable	G3 — Final period grade (0–20 scale)

Key Variables Used:

studytime   — Weekly study time (1=<2h, 2=2-5h, 3=5-10h, 4=>10h)
failures    — Number of past class failures (0–4)
absences    — Number of school absences
Medu/Fedu   — Mother's / Father's education level (0–4)
Dalc/Walc   — Workday / Weekend alcohol consumption (1–5)
goout       — Going out with friends (1–5)
sex         — Student gender (M/F)
G1, G2, G3  — Period grades 1, 2, 3 (0–20)

---

🛠 Tech Stack & Tools

Layer	Tool	Purpose
🐍 Language	Python 3.11	Core analysis language
📊 Visualisation	Matplotlib 3.8	Chart generation & PDF export
🧮 Data	Pandas 2.1 · NumPy 1.26	Data loading, cleaning, aggregation
📐 Statistics	SciPy 1.11	ANOVA, t-tests, Pearson correlation
🌐 Dashboard	Streamlit 1.29	Interactive web interface
🖼 Image	Pillow 10.1	PNG loading in Streamlit

---

📁 Folder Structure

student-performance-analysis/
│
├── 📄 app.py                          ← Streamlit dashboard (entry point)
├── 📄 requirements.txt                ← Python dependencies
├── 📄 README.md                       ← You are here
│
├── 📂 data/
│   ├── student-mat.csv                ← Raw Mathematics dataset (Kaggle/UCI)
│   ├── student-por.csv                ← Raw Portuguese dataset (Kaggle/UCI)
│   └── student-mat-clean.csv          ← Cleaned & preprocessed working dataset
│
├── 📂 python/
│   ├── 01_preprocessing.py            ← Data loading, cleaning, distributions
│   ├── 02_studytime_analysis.py       ← Study time vs G3 analysis
│   ├── 03_parental_education.py       ← Parental education impact
│   ├── 04_gender_analysis.py          ← Gender-based performance analysis
│   ├── 05_correlation.py              ← Heatmap & pairplot
│   ├── 06_lifestyle.py                ← Alcohol & social lifestyle analysis
│   └── 07_hypothesis_testing.py       ← ANOVA & t-test statistical testing
│
├── 📂 R/
│   └── analysis.Rmd                   ← Full R Markdown analysis document
│
├── 📂 outputs/
│   ├── 📂 charts/                     ← All generated PNG + PDF chart exports
│   └── 📂 reports/                    ← Final analysis summary reports
│
└── 📂 docs/                           ← Project documents (FRD, TD, TC, MD)

---

🔹 Module Breakdown

Module 01 — Data Preprocessing & Cleaning

Script: python/01_preprocessing.py

Loads the cleaned dataset, inspects data types, identifies null values, and generates exploratory distribution charts for all grade columns and categorical features.

Charts Generated:

• Grade distribution histograms (G1, G2, G3)
• Missing value bar chart
• Categorical column value counts

---

Module 02 — Study Time vs Final Grade

Script: python/02_studytime_analysis.py

Analyses how weekly study hours (4-level scale) relate to G3. Includes regression analysis and group comparisons.

Charts Generated:

Key Finding: Students studying 5–10 hrs/week score on average 2–3 points higher in G3 than those studying less than 2 hrs.

---

Module 03 — Parental Education Impact

Script: python/03_parental_education.py

Compares mean G3 across all 5 education levels for both mother (Medu) and father (Fedu).

Charts Generated:

Key Finding: A clear positive gradient exists — students with university-educated parents score 2–4 points higher in G3.

---

Module 04 — Gender-Based Performance

Script: python/04_gender_analysis.py

Visualises grade distributions, progression (G1→G2→G3), and tests gender differences using independent t-tests with effect size (Cohen's d).

Charts Generated:

---

Module 05 — Correlation Analysis

Script: python/05_correlation.py

Computes Pearson correlation matrix across studytime, failures, absences, G1, G2, G3 and renders annotated visualisations.

Charts Generated:

Heatmap	Pairplot	G3 Predictors

Key Finding: G2 (r > 0.90) and failures (r ≈ −0.36) are the strongest predictors of G3.

---

Module 06 — Lifestyle Impact Analysis

Script: python/06_lifestyle.py

Two sub-sections: Alcohol consumption (Dalc, Walc) and Social activity (goout) vs G3.

Charts Generated:

Alcohol vs Grades:

Bar Chart	Box Plot	Scatter

Social Activity vs Grades:

Goout Bar	Goout Scatter	Combined Heatmap

Key Finding: Students with Very High weekend alcohol + Very Often going out represent the worst-performing cohort.

---

Module 07 — Hypothesis Testing

Script: python/07_hypothesis_testing.py

Statistically validates two core findings using formal hypothesis testing.

H₁ — Does study time significantly affect grades? (One-Way ANOVA)

Box Plot	CI Bars

• Includes Levene's test for equal variances
• Bonferroni-corrected pairwise t-tests for post-hoc analysis
• Significance brackets annotated directly on charts

H₂ — Is there a significant gender difference in G3? (Independent t-test)

• Welch's t-test (handles unequal variances)
• Cohen's d for practical effect size
• Significance notation: * p<.05 · ** p<.01 · *** p<.001

---

🌐 Streamlit Dashboard

The interactive dashboard connects all 7 modules into a single colour-rich app with dedicated pages per module and a 4-tab Showcase Questions section.

▶️ Quick Start

# 1. Clone the repository
git clone https://github.com/YOUR-USERNAME/student-performance-analysis.git
cd student-performance-analysis

# 2. Install dependencies
pip install -r requirements.txt

# 3. Run analysis scripts first (to generate charts)
cd python
python 01_preprocessing.py
python 02_studytime_analysis.py
python 03_parental_education.py
python 04_gender_analysis.py
python 05_correlation.py
python 06_lifestyle.py
python 07_hypothesis_testing.py

# 4. Launch the dashboard
cd ..
streamlit run app.py

The dashboard will open at http://localhost:8501

📱 Dashboard Pages

🏠 Overview              → KPI cards + module grid + showcase preview
🔹 01 Preprocessing      → Grade distributions + categorical charts
🔹 02 Study Time         → Study hours vs G3 analysis
🔹 03 Parental Education → Parent education level vs grades
🔹 04 Gender Analysis    → Gender performance + progression
🔹 05 Correlation        → Heatmap + pairplot + predictor bar
🔹 06 Lifestyle          → Alcohol + social activity charts
🔹 07 Hypothesis Testing → ANOVA + t-test results + annotations
💡 Showcase Questions    → 4-tab policy insight dashboard

---

💡 Showcase Questions

The dashboard's Showcase section provides data-backed answers to four policy questions:

🧑‍👩‍👧 Support Systems For students with low parental education Students with no-education parents score 2–4 points lower in G3. Recommended interventions: Academic mentorship pairing Parent engagement workshops Free tutoring & resource access	📉 Academic Intervention For students with poor academic trends Failures & G2 score are the strongest predictors. Recommended strategies: Early Warning System (flag G1/G2 < 8) Repeat-failure remedial tracks Personalised grade improvement goals
🍺 Lifestyle Awareness Programs on lifestyle's academic impact Very High Walc students score 1.5–3 points lower. Recommended programs: Data-backed campus talks Student self-assessment lifestyle quiz Healthy alternative activity promotion	📚 Study Habit Strategies Improving student study habits Moving from <2 hrs to 5–10 hrs/week = +2–3 grade points (ANOVA confirmed). Strategies: Structured weekly timetables Peer study group sessions Gamified learning with point systems

---

📊 Results Summary

Finding	Statistical Test	Result
Study time → G3	One-Way ANOVA	✅ Significant (p < 0.05)
Gender → G3	Independent t-test (Welch)	Depends on dataset
G2 → G3 correlation	Pearson r	r > 0.90 (very strong)
Failures → G3	Pearson r	r ≈ −0.36 (moderate negative)
Alcohol → G3	Pearson r + regression	Negative trend
Parental education → G3	Group mean comparison	Clear positive gradient

---

🚀 Quick Start

# Install all dependencies
pip install -r requirements.txt

# Run a single module
python python/05_correlation.py

# Launch the full dashboard
streamlit run app.py

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Built with 🎓 Python · Pandas · Matplotlib · SciPy · Streamlit

⭐ If you found this project useful, consider giving it a star! ⭐