🎓 Student Performance Predictor

A machine learning project that predicts whether a student will pass or fail based on personal, social, and academic factors. Built as a capstone for the Fundamentals of AI and ML course.

---

📌 Problem Statement

In many educational institutions, students who are at risk of failing are identified only after it is too late to intervene. This project builds an early-warning classifier using supervised learning that can flag at-risk students based on observable characteristics, enabling teachers and counsellors to provide timely support.

---

🧠 ML Concepts Used

Concept	Application
Supervised Learning	Binary classification (Pass / Fail)
Decision Tree Classifier	Main model — interpretable and visual
Train/Test Split	80/20 stratified split
Model Evaluation	Accuracy, Precision, Recall, F1-Score, Confusion Matrix
Feature Importance	Identifying the strongest predictors of student outcomes

---

📂 Project Structure

student-performance-predictor/
│
├── student_performance_predictor.py   # Main ML script
├── student-mat.csv                    # Dataset
├── results.png                        # Output: confusion matrix + feature importance chart
├── README.md                          # This file
└── report/
    └── Project_Report.pdf             # Full project report

---

📊 Dataset

Source: UCI Machine Learning Repository — Student Performance Dataset

• File needed: student-mat.csv (Mathematics course data)
• Students: 395
• Features: 33 attributes covering demographics, family background, study habits, and grades
• Target: Final grade G3 → converted to Pass (≥10) or Fail (<10)

---

⚙️ Setup & Installation

1. Clone the repository

git clone https://github.com/<your-username>/student-performance-predictor.git
cd student-performance-predictor

2. Install dependencies

pip install pandas numpy scikit-learn matplotlib seaborn

Python 3.8 or higher is recommended.

3. Add the dataset

Download student-mat.csv from the UCI link above and place it in the project root folder.

4. Run the script

python student_performance_predictor.py

---

📈 Sample Output

After running, you will see:

• Accuracy score and full classification report in the terminal
• A saved image results.png containing:
  • Confusion Matrix (actual vs predicted outcomes)
  • Feature Importance bar chart (which factors matter most)
• Human-readable decision rules extracted from the tree
• A demo prediction for a sample student profile

Example terminal output:

Accuracy : 91.14%

Classification Report:
              precision    recall  f1-score   support
        Fail       0.79      0.73      0.76        15
        Pass       0.93      0.95      0.94        64

    accuracy                           0.91        79

---

🔑 Key Findings

• G2 (second-period grade) is the strongest predictor of final outcome — prior performance is highly predictive.
• Number of past failures is the second most important feature.
• Students who aspire to higher education (higher = yes) significantly outperform those who do not.
• Lifestyle factors like goout and Walc (weekend alcohol consumption) have a measurable negative impact.

---

🛠️ Dependencies

Package	Version
Python	≥ 3.8
pandas	≥ 1.3
numpy	≥ 1.21
scikit-learn	≥ 1.0
matplotlib	≥ 3.4
seaborn	≥ 0.11

Install all at once:

pip install pandas numpy scikit-learn matplotlib seaborn

---

🔮 Future Improvements

• Add a web interface using Streamlit so teachers can input student details and get a prediction
• Try Random Forest or Gradient Boosting for better accuracy
• Incorporate the Portuguese language course dataset (student-por.csv) for broader generalisation
• Use SHAP values for more detailed model explainability

---

📄 License

This project is submitted as part of an academic course evaluation. Dataset credit: P. Cortez and A. Silva, University of Minho, Portugal.

---

👤 Author

Shivam Rathore
Course: Fundamentals of AI and ML