README.md

metadata

title: Calorie Expenditure Predictor
emoji: 🔥
colorFrom: purple
colorTo: red
sdk: streamlit
sdk_version: 1.31.0
python_version: '3.10'
app_file: app.py
pinned: false
license: mit

Calorie Expenditure Prediction - Kaggle Playground Series S5E5

This repository contains a machine learning application deployed to Hugging Face Spaces that predicts calories burned during physical activity. This project was developed as part of a "Become a Pro" Data Science challenge.

Project Overview

The goal of this project is to provide high-precision caloric expenditure estimates based on physiological and exercise-related metrics. The model was trained on a dataset of 750,000 records, focusing on memory efficiency and non-linear relationship modeling.

Technical Details

• Model: XGBRegressor (Extreme Gradient Boosting)
• Performance:
  • R-Squared: 0.9962
  • RMSE: 3.80
• Environment: Python 3.10
• Key Techniques:
  • Manual Categorical Mapping (Female: 1, Male: 0) to optimize memory.
  • Histogram-based tree method (tree_method='hist') for handling large-scale tabular data within 8GB RAM constraints.

Features Used

• Sex (Encoded)
• Age
• Height
• Weight
• Duration
• Heart_Rate
• Body_Temp

Local Setup

To run this project locally on VS Code with Python 3.10:

1. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use: venv\Scripts\activate