This article is cited in 1 paper

Bayesian classification of serial multivariate data

M. P. Krivenko

Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The analysis of the results of observations of a group of objects during the time when these objects can change any of their significant characteristics is considered. The goal is to describe the changes and identify the factors that determine them. Appropriate methods are known as longitudinal. The article proposes a different approach when a series of multivariable characteristics of an object makes up a single vector of the observed values. By increasing the dimension of the data, it is possible to obtain a single picture of the description of objects and to formalize the construction of a data model. To demonstrate the essence of the approach and illustrate the emerging possibilities of data analysis, the problem of early cancer diagnosis using a prostate-specific antigen biomarker is considered. It was revealed that a multivariable approach to the analysis of a series of analyzes leads to an increase in the accuracy of diagnosis.

Keywords: serial data classification, longitudinal analysis, consolidation approach, mixture of probabilistic principal component analyzers.

Received: 09.01.2020

DOI: 10.14357/08696527200103