B1050
Title: Individuality-based fuzzy cluster-scaled principal component analysis for high-dimension low-sample data
Authors: Mika Sato-Ilic - University of Tsukuba (Japan) [presenting]
Abstract: A fuzzy clustering-based Principal Component Analysis (F-PCA) is presented by considering the individuality of subjects for high-dimension, low-sample size (HDLSS) data. Analyzing HDLSS data, including the difference of subjects, is useful as an implementation of a custom-made system for healthcare considering the individual history of daily activities. For example, if we observe data over subjects by sensors worn on the body during activities, analyzing high dimensional times with a low number of sensor measurements over the subjects is useful for implementing the custom-made system for healthcare. A simultaneous analysis is necessary for obtaining the well-classified result among different subjects and the efficient reduction of high dimensional data with sufficient explainability of the original data. Without external information on the difference between the subjects, we need to capture the difference of subjects from the single target original data. Since the fundamental idea of F-PCA is the inclusion of the weights measured by the fuzzy cluster scale commonly obtained over the subjects to the covariance of variables of the original data, the proposed F-PCA is adaptable to this analysis. The proposed F-PCA is shown to have a better performance with a comparison of the results of ordinary PCA by several numerical examples.