Tri-Axial Accelerometer Data-Driven Senior Activity Recognition: Multi-Class Machine Learning Approaches

As global aging intensifies, the demand for health monitoring and services for the elderly has significantly increased. Human Activity Recognition (HAR) technology is considered a key solution to addressing these challenges. This study aims to establish a standard paradigm for HAR using the HAR70+ dataset from the University of California, Irvine Machine Learning Repository. The dataset is extensive, and various preprocessing methods are employed. Cross-validation is used to split the training and validation sets, and multiple metrics are applied to assess model performance. The models utilized in this study include statistical, machine learning, and deep learning algorithms such as logistic regression, random forests, and long short-term memory (LSTM) networks. A comparative analysis is conducted by evaluating overall prediction accuracy, label-wise accuracy, variable combination accuracy, and loss curves. Those analyses help identify the optimal algorithm for the dataset and explore the best variable combinations. The preprocessing workflows, comparison methods, and analytical approaches adopted in this study contribute to creating a standard paradigm for HAR technology.