🔥 Calorie Prediction Microservice App

This project is a simple, interactive web application that predicts calories burned during exercise, based on user input as following: age, weight, height,duration of exercise, heart rate, and current body temperature.

It follows a microservice-based architecture using:

• Streamlit as the frontend UI
• FastAPI as the backend prediction API

---

🧠 What It Does

• Collects user details (like age, height, weight, duration of exercise, etc.) through a friendly web interface.
• Sends that data to a FastAPI microservice.
• Applies feature engineering and a trained machine learning model to predict calories burned.
• Returns the prediction to the frontend and displays it.

---

---

🚀 How to Run It Locally

Make sure you have Python 3.8+ installed. Then:

1. Clone the repo

git clone https://github.com/sfarrukhm/predict-calories.git
cd predict-calories

2. Install dependencies

Create and activate a virtual environment first:

python -m venv venv
source venv/bin/activate   # or venv\Scripts\activate on Windows

# Then install dependencies
pip install -r requirements.txt

3. Start the FastAPI backend

cd predictor_service
uvicorn main:app --reload

This will start the API server at http://localhost:8000

4. Start the Streamlit frontend

cd frontend
streamlit run app.py

📚 Inspiration & Data Source

This project is based on the dataset provided in the Kaggle Playground Series - Season 5, Episode 5, which involves predicting calories burned during exercise.

The feature engineering and modeling techniques can be seen in following notebook:

📓 KMeans Clustering-Based FE Improves CV & LB Scores