🎓 Student Performance Predictor

📌 Overview

This project focuses on predicting student academic performance based on demographic and lifestyle attributes.
Using a dataset with factors such as gender, ethnicity, parental education, lunch type, and test preparation, the goal is to forecast student math, reading, and writing scores through machine learning regression models.

The project evaluates multiple models and benchmarks their predictive power to identify the most reliable approach.

---

📊 Dataset Description

The dataset includes the following columns:

• gender – Student gender
• race_ethnicity – Student’s ethnic background
• parental_level_of_education – Parent’s highest education level
  • Options: associate's degree, bachelor's degree, high school, master's degree, some college, some high school
• lunch – Type of lunch (free/reduced, standard)
• test_preparation_course – Completion status of prep course
• math_score – Math exam score
• reading_score – Reading exam score
• writing_score – Writing exam score

---

🚀 Models Trained & Results

🔹 Linear Regression

• Train: RMSE 5.34 | MAE 4.27 | R² 0.8735
• Test: RMSE 5.42 | MAE 4.22 | R² 0.8792 ✅

🔹 Lasso

• Train: RMSE 6.59 | R² 0.8071
• Test: RMSE 6.52 | R² 0.8253

🔹 Ridge

• Train: RMSE 5.32 | R² 0.8743
• Test: RMSE 5.39 | R² 0.8806 ✅ (Best among linear models)

🔹 K-Neighbors Regressor

• Train: RMSE 5.71 | R² 0.8554
• Test: RMSE 7.26 | R² 0.7835

🔹 Decision Tree

• Train: RMSE 0.27 | R² 0.9997 (Overfitting)
• Test: RMSE 7.35 | R² 0.7778

🔹 Random Forest Regressor

• Train: RMSE 2.28 | R² 0.9769
• Test: RMSE 5.99 | R² 0.8527

🔹 XGBRegressor

• Train: RMSE 1.01 | R² 0.9955
• Test: RMSE 6.47 | R² 0.8278

🔹 CatBoost Regressor

• Train: RMSE 3.04 | R² 0.9589
• Test: RMSE 6.01 | R² 0.8516

🔹 AdaBoost Regressor

• Train: RMSE 5.84 | R² 0.8485
• Test: RMSE 6.21 | R² 0.8413

📌 Observation:

• Ridge Regression shows the most stable and best generalization with R² ≈ 0.88.
• Tree-based models like Random Forest, CatBoost, and AdaBoost also perform competitively.
• Decision Tree & XGBoost exhibit signs of overfitting.

---

🛠️ Tech Stack

• Python
• Flask (for deployment)
• NumPy, Pandas, Seaborn, Matplotlib
• Scikit-learn
• CatBoost, XGBoost
• Dill (model serialization)

---

📂 Project Structure

ML-Student_Performance_Predictior/
├── artifacts/ # Saved models, metrics, reports, etc.
├── catboost_info/ # CatBoost training logs and artifacts
├── notebook/ # Exploratory data analysis & experiments
├── src/ # Source code for data processing, model training, inference
├── templates/ # HTML templates for Flask app frontend
├── app.py # Main Flask application script
├── requirements.txt # Project dependencies
├── setup.py # Setup configuration for installation
├── README.md # Project documentation
└── .gitignore # Files/folders to ignore in Git

🚀 How to Run

Follow these steps to set up and run the project locally:

1. Clone the repository

   git clone https://github.com/DipeanDas/ML-Student_Performance_Predictior.git
   cd ML-Student_Performance_Predictior

2. Create and activate a virtual environment

   python -m venv venv
   source venv/bin/activate    # On Linux/Mac
   venv\Scripts\activate       # On Windows
   

3. Install dependencies

   pip install -r requirements.txt
   

4. Run the Flask app

   python app.py
   

Project By

Dipean Dasgupta
Data Analyst & AIML Enthusiast