A0687
Title: Sparse clustering for multi-dimensional functional data using Frechet distance: Methods and applications in medical data
Authors: Soon-Sun Kwon - Ajou University (Korea, South) [presenting]
Abstract: Novel clustering methodologies are presented for multi-dimensional functional data that account for asynchronous measurements and varying variable contributions. Building on the Frechet distance, two scalable algorithms are proposed: a multi-dimensional extension of K-means and a sparse variant that incorporates functional variable selection. These methods effectively group irregularly observed trajectories and identify the most influential variables in the clustering process. The performance of the approaches are validated through extensive simulation studies and apply the sparse clustering technique to longitudinal thyroid cancer data from South Korea. The analysis reveals clinically meaningful patient clusters and highlights the role of specific biomarkers in disease prognosis. This framework offers a powerful tool for high-dimensional trajectory analysis, especially in biomedical settings.
