Мethodology of information monitoring and diagnostics of objects represented by quantitative estimates based on cluster analysis

Abstract

The paper discusses the methodological foundations of informational diagnostics on the base of cluster analysis
for the objects represented by quantitative estimates. The literature review showed that the application of cluster analysis in some
cases was successful; also, the theory of cluster analysis is well developed, and the properties of methods and distance measures are
studied, which indicates the appropriateness of using the cluster analysis apparatus. Therefore, the development of a general
methodology to diagnose any objects represented by quantitative estimates is a topical task. The purpose of this work is to develop
methodological bases for determining diagnostic states and behavioral patterns for objects represented by quantitative estimates on
the base of cluster analysis. Because of informational diagnostics is a targeted activity on the assessment of object state based on a
dynamic information model, the model of a diagnosis object is discussed first. We examine the lifecycle of instances of diagnosis
objects that are described by a plurality of parameters whose values are determined by a time slice along the lifeline of the instance.
It is shown that a different number of measured values characterize each state of the diagnosis object. There are identified
characteristics that should be analyzed to indicate a threat to the instance and the need for supportive procedures to prevent
premature interruption of an instance's lifecycle. Experts should carry out the formalization of conditions for termination of the life
cycle of the diagnosis object and formation of the list of supporting procedures. Because the quality of any information technology
depends on the input data quality, a procedure for the analysis of diagnostic characters is developed. In order to start the diagnosis
as early as possible and apply the available data as fully as possible, the methodologies for one-, two- and N-step diagnosis are
developed. All procedures used cluster order. Transition patterns are defined for the two-step diagnosis, as well as trend patterns are
defined for the N-step diagnosis. Transition patterns allow diagnosing the improvement, worsening, or stability of the diagnosis
object state. The procedure for the diagnostic characters analysis and the methodologies of diagnosis is new scientific results. The
application of the developed methodologies is demonstrated in the example of diagnosing students' success. In this case, the
curriculum provides the domain model. Examples of diagnosing states and behavior, as well as identifying recommended reactions,
are provided. For one-step diagnostics, the presence of the influence of the latent factor and the diagnostic signs that show
significant instability are investigated. For one- and two-step diagnostics, the conditions for forming a risk segment are provided