Skip to content

Latest commit

 

History

History
92 lines (72 loc) · 4.37 KB

README.md

File metadata and controls

92 lines (72 loc) · 4.37 KB

SmartFit AI

[Streamlit App] (https://smartfit-ai.streamlit.app/)

License: MIT](https://opensource.org/licenses/MIT)

SmartFit AI is a comprehensive machine-learning-powered system that analyzes workout patterns, dietary habits, and health indicators to deliver personalized insights, predictions, and recommendations. It combines supervised, unsupervised, and deep-learning techniques to model calorie burn, cluster fitness profiles, and suggest optimal workouts and diet plans._

##Dashboard Preview image

Key Highlights

  • Predict calories burned, BMI, and fat percentage
  • Identify user fitness/diet archetypes using clustering
  • Build neural-network models for health profiling
  • Recommend personalized diet and workout routines
  • Interactive Streamlit dashboard for live exploration

Table of Contents

Overview

SmartFit AI leverages a dataset of user fitness metrics to provide actionable insights. Using techniques like PCA for dimensionality reduction, K-Means clustering for profile segmentation, and neural networks for predictions, it helps users optimize their health journeys. The interactive Streamlit app visualizes data, predicts outcomes, and generates recommendations in real-time.

Features

  • Predictions: Real-time calorie burn, BMI, body fat analysis, and workout impact projections.
  • Clustering- Unsupervised learning to group users into 5 fitness archetypes (Elite Athletes, Strength Builders, Enthusiasts, Beginners, Health Focus).
  • Recommendations: AI-driven workout schedules and diet plans tailored to user stats, goals, and equipment.
  • Visualizations: Radar charts, pie charts, heatmaps, scatter plots, and more for intuitive data exploration.
  • Data Analysis: Correlation heatmaps, PCA visualizations, and distribution analyses.

Dashboard Sections

The Streamlit app is organized into intuitive sections:

  1. ** Dashboard**

    • System overview with key metrics.
    • Cluster distribution visualization.
    • Calorie burn by workout type.
    • BMI distribution analysis.

  2. ** Predictions**

    • Calorie Burn Calculator: Real-time prediction based on intensity, duration, and user stats.
    • BMI & Body Fat Analyzer: With visual gauge charts.
    • Workout Impact Predictor: Project weight changes over time with timeline charts.

  3. ** Fitness Profiles**

    • 5 Fitness archetypes (Elite Athletes, Strength Builders, Enthusiasts, Beginners, Health Focus).
    • Interactive profile matching based on user input.
    • Radar charts showing fitness attributes.
    • Detailed cluster characteristics.

  4. ** Diet Planner**

    • Personalized macronutrient calculations.
    • Sample meal plans for different goals.
    • Macronutrient distribution pie charts.
    • Weekly shopping lists.

  5. ** Workout Recommender**

    • AI-generated workout schedules.
    • Detailed strength, cardio, and recovery sessions.
    • Progress tracking metrics.
    • Customized based on experience and equipment.

  6. ** Data Explorer**

    • Correlation heatmaps.
    • Interactive distribution visualizations.
    • 2D/3D scatter plots.
    • PCA cluster visualization with explained variance.

Dataset

  • Shape: (20,000, 62) – 20,000 rows with 62 features.
  • Key Columns and Datatypes:
    • age: float64
    • gender: object
    • weight_kg: float64
    • height_m: float64
    • max_bpm: float64
    • ... (additional features like pct_fats: float64, difficulty_level_enc: int32, cluster: int32, pca1: float64, pca2: float64)
  • Missing Values (%): 0.0 across all columns (e.g., age: 0.0, protein_per_kg: 0.0, sets: 0.0, benefit: 0.0, sodium_mg: 0.0, cholesterol_mg: 0.0, serving_size_g: 0.0,0.0).
  • The dataset is fully clean with no missing values, making it ideal for direct modeling.

Model Performance

  • Clustering: K-Means applied on PCA-reduced features (pca1, pca2).
  • Clustering: K-Means applied on PCA-reduced features (pca1, pca2) to identify 5 fitness archetypes. Explained variance from PCA can be visualized in the app.
  • Supervised Models (e.g., Neural Networks for calorie burn, BMI, and body fat prediction)