Student Math Score Predictor

An end-to-end Machine Learning project that predicts a student's math score based on demographic and academic features. Built with modular code, containerized with Docker, and deployed on AWS ECS via an automated GitHub Actions CI/CD pipeline.

---

Problem Statement

Predict a student's math exam score given inputs like gender, parental education level, lunch type, test preparation course completion, and reading/writing scores.

Target variable: math_score (continuous — regression problem)

---

App Preview

---

Architecture

Raw Data → Data Ingestion → Data Transformation → Model Training
                                                        ↓
User Input → Flask Web App → Prediction Pipeline → Predicted Score
                                                        ↑
                                              Model + Preprocessor (artifacts/)

---

Project Structure

├── .aws/                        # ECS task definition
├── .github/workflows/aws.yml    # GitHub Actions CI/CD pipeline
├── src/
│   ├── components/
│   │   ├── data_ingestion.py
│   │   ├── data_transformation.py
│   │   └── model_trainer.py
│   ├── pipeline/
│   │   ├── train_pipeline.py
│   │   └── predict_pipeline.py
│   ├── exception.py             # Custom exception handler
│   ├── logger.py                # Centralized logging
│   └── utils.py                 # save/load object, model evaluation
├── artifacts/  
│   ├── data.csv
│   ├── train.csv
│   ├── test.csv
│   ├── preprocessor.pkl
│   └── model.pkl               
├── notebook/ 
│   └── data/   
│       └── stud.csv             # student data
├── templates/                   # Flask app templates
│   ├── home.html
│   └── index.html              
├── app.py                       # Flask application
├── Dockerfile
├── requirements.txt
└── setup.py

---

Input Features

Feature	Type	Example
Gender	Categorical	male / female
Race/Ethnicity	Categorical	group A–E
Parental Education	Categorical	bachelor's degree
Lunch	Categorical	standard / free-reduced
Test Prep Course	Categorical	completed / none
Reading Score	Numeric	72
Writing Score	Numeric	68

---

Models Evaluated

Trained and compared using GridSearchCV (3-fold CV), scored by R²:

• Linear Regression
• Random Forest
• Decision Tree
• Gradient Boosting
• AdaBoost

Best model and preprocessor are saved to artifacts/ for inference.

---

Results

Best Model	R² Score
Linear Regression	0.88

---

Quickstart

Local Setup

git clone https://github.com/Priyesh-DS-Code/ML-AWS-CI-CD-Project.git
cd ML-AWS-CI-CD-Project

python -m venv venv && source venv/bin/activate   # Windows: venv\Scripts\activate
pip install -r requirements.txt

python src/pipeline/train_pipeline.py   # trains model, saves artifacts
python app.py                           # starts Flask on http://localhost:5000

Docker

docker build -t student-score-predictor .
docker run -p 5000:5000 student-score-predictor

---

CI/CD — GitHub Actions + AWS

On every push to main:

1. Docker image is built
2. Pushed to AWS ECR
3. Deployed to AWS ECS (Fargate)

Required GitHub Secrets:

AWS_ACCESS_KEY_ID
AWS_SECRET_ACCESS_KEY
AWS_REGION
ECR_REPOSITORY
ECS_CLUSTER
ECS_SERVICE

---

Tech Stack

Python · Scikit-learn · Flask · Docker · AWS ECR/ECS · GitHub Actions

---

Connect

Priyesh · GitHub