🎓 Student Performance Prediction - MLOps Zoomcamp Project

This project is part of the MLOps Zoomcamp and demonstrates a complete MLOps workflow for building, deploying, and monitoring a machine learning model.

The goal is to predict the final grade (G3) of students in a math course using a dataset of demographic, academic, and social features.

---

📊 Dataset

The dataset comes from the UCI Machine Learning Repository and contains various student attributes:

• Demographic: age, sex, address type
• Social: family size, parental status, romantic relationships
• Academic: study time, failures, previous grades (G1, G2), support classes
• Target: G3 - final math grade (0–20)

Data source: UCI Student Performance Data Set

---

⚙️ Project Structure

├── data/ # Raw and processed data

├── notebooks/ # EDA, experimentation and training

└── eda.ipynb # Exploratory Data Analysis

└── eda.py # Exploratory Data Analysis Script

└── preprocessing_and_modeling_with_mlflow.ipynb

└── preprocessing_and_modeling_with_mlflow.py

├── serving/ # Model serving with FastAPI

└── main.py # FastAPI app for model inference

├── requirements.txt # Python dependencies

├── README.md # Project overview

---

🚀 MLOps Stack

• Experiment Tracking: MLflow
• Model Training: scikit-learn, pandas
• Model Serving: FastAPI

---

🛠️ Setup Instructions

Clone the repo

Clone https://github.com/elvisiraguha/student-performance-mlops

git clone https://github.com/elvisiraguha/student-performance-mlops.git
cd student-performance-mlops

Create virtual environment

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Run training

python notebooks/preprocessing_and_modeling_with_mlflow.py

3. 🚀 Launch MLflow UI

mlflow ui

Visit: http://localhost:5000

• View experiment runs

• Compare metrics (MAE, RMSE, R²)

• Manage registered models (Staging/Production)

🧠 Model Serving via FastAPI

We serve the registered model using FastAPI + MLflow model loading.

Run the API server

uvicorn serving.main:app --reload

📈 Sample Prediction Request

curl -X POST http://localhost:8000/predict \
  -H "Content-Type: application/json" \
  -d '{
        "school": "GP", "sex": "F", "age": 17, "address": "U", "famsize": "GT3",
        "Pstatus": "T", "Medu": 4, "Fedu": 4, "Mjob": "teacher", "Fjob": "services",
        "reason": "course", "guardian": "mother", "traveltime": 1, "studytime": 2,
        "failures": 0, "schoolsup": 0, "famsup": 1, "paid": 0, "activities": 1,
        "nursery": 1, "higher": 1, "internet": 1, "romantic": 0, "famrel": 5,
        "freetime": 3, "goout": 3, "Dalc": 1, "Walc": 2, "health": 5, "absences": 3,
        "G1": 14, "G2": 15
      }'

✅ Features Completed

• Data cleaning and preprocessing

• EDA and feature exploration

• Model training and evaluation

• MLflow experiment logging

• Model registration to Model Registry

• Model serving via FastAPI

📦 Next Steps

• Dockerize training and serving

• CI/CD for model updates

• Monitoring with Prometheus + Grafana

• Deploy to cloud (e.g., GCP, AWS, or Azure)

🧑‍💻 Author

Elvis Iraguha – @elvisiraguha

📄 License

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