Gym Member Exercise Prediction

This project demonstrates machine learning models to analyze and predict gym member exercise patterns using Python.

Project Structure

• Prediction_using_function.ipynb - Implementation using modular functions
• Prediction_without_function.ipynb - Direct implementation without functions
• Prediction_with_user_input.ipynb - Interactive implementation allowing user input
• Calories_burned_prediction_optimized.ipynb - Optimized implementation with improved model architecture

Dependencies

• NumPy
• Pandas
• Matplotlib
• PyTorch
• Scikit-learn
• StandardScaler

Dataset

The project uses a gym members exercise tracking dataset accessed from: https://raw.githubusercontent.com/sayande01/Kaggle_Notebooks/refs/heads/main/gym_members_exercise_tracking.csv

Features

• Data preprocessing and standardization
• Linear regression model implementation using PyTorch
• Model evaluation metrics (MSE, MAE, R-squared)
• Data visualization of predictions vs actual values
• User input interface for real-time predictions
• Optimized model architecture with improved performance
• Learning rate tuning and epoch optimization
• Enhanced data visualization for model training progress

Model Features

The model predicts calories burned based on:

• Age
• Weight (kg)
• Height (m)
• Max BPM
• Average BPM
• Resting BPM
• Session Duration (hours)
• Fat Percentage
• Water Intake (liters)
• Workout Frequency (days/week)
• Experience Level
• BMI
• Gender
• Workout Type

Usage

1. Open notebooks in Jupyter/Google Colab
2. Run cells sequentially to:
   • Load and preprocess data
   • Train the prediction model
   • Evaluate model performance
   • Make predictions
3. Use Prediction_with_user_input.ipynb for custom value predictions
4. Use Calories_burned_prediction_optimized.ipynb for best performance

Implementations

The project provides four different approaches:

1. Modular approach with functions for better code organization
2. Direct implementation for simpler understanding
3. Interactive implementation with user input functionality
4. Optimized implementation with enhanced model architecture and performance

Model Performance

• Mean Squared Error (MSE): ~1560 (improved from ~2100)
• Mean Absolute Error (MAE): ~30 (improved from ~36)
• R-squared (R²): ~0.979 (improved from ~0.97)