LightGBM for Human Activity Recognition Using Wearable Sensors

Abstract

With the popularization of information technology, a variety of embedded sensors have quietly entered our side, through these smart devices, can get our human body activity signals, so its life services, health monitoring, sports training and other fields play a vital role, and how to identify these signals for the focus of the study. In view of the development in the field of machine learning, especially the excellent performance of LightGBM (Light Gradient Boosting Machine) algorithm in handling large datasets and high-dimensional features, this study used HAR data collected based on smartphones and applied LightGBM algorithm on a publicly available dataset to achieve 94.40% accuracy rate. To enhance the persuasiveness, we also chose KNN (K-Nearest Neighbors) and Decision Tree as comparison experiments.