WISDM Health Monitoring

This project aims to compare the performance of various machine learning models for the WISDM activity tracking dataset. Tracking human activity correctly would allow for better health monitoring systems which would lead to healthier lifestyle. The dataset comprises of information from 51 subjects who performed 18 activities for 3 minutes each which we then grouped into 3 main classes, ”Non Hand Movements”, ”Hand Oriented General Movements” and ”Hand Oriented Eating Movements”. The data is collected from their phone’s accelerometer and gyroscope as well as smart watch’s accelerometer and gyroscope and is sampled 20 times per second (20 Hz). In this project we have used a dataset which was derived from the original dataset and has statistical features computed from the time series data. In this work, we compare models like Logistic Regression, Decision Trees, Random Forest, XGboost, AdaBoost, AutoML and Artificial Neural Networks.