Human Activity Recognition using Ensemble Learning

This project uses statistical features from tri-axial acceleration and gyroscopic signals to classify human activities such as walking, sitting, and laying using machine learning models.

Dataset

• UCI Human Activity Recognition Dataset
• Dataset Link

Features

• 561 pre-extracted statistical features (mean, std, etc.)
• Activities classified: Walking, Walking Upstairs, Walking Downstairs, Sitting, Standing, Laying

Models Compared

• Random Forest
• XGBoost
• LightGBM ✅ (Best Accuracy + Speed)
• Stacking Classifier
• Voting Classifier
• PCA + XGBoost

Special Feature

Detects sudden risky transitions (e.g., Walking → Laying) that could indicate a fall.

How to Run

1. Install dependencies

pip install -r requirements.txt