Acitivity Recognition Using Single Chest Mounted Accelerometer

Author: Nischay Thapa

Physical activity recognition is a comprehensive study that intends to identify a person's actions based on sensor data. In this project, activity recognition is expressed as a multi class classification problem where the task is to predict seven different activities. We evaluated model's performance based on macro average f1-score and found that 5 neighbour classifers performs well in identifying those activities.

Data File

• All data files are retrieved from UCI Machine Learning Repository and are stored in Activity Recognition from Single Chest Mounted Accelerometer

• Preprocessed data are stored as preprocessed_data.csv file

• For detail experiements, refer to this notebook.

Data Exploration

Acceleration recorded on x-axis

Acceleration recorded on y-axis

Acceleration recorded on z-axis

Target Label Distribution

Test Set Result

Classification Report

Confusion Matrix