АННОТАЦИЯ В статье предлагается метод ко-кластеризации стохастических данных многомерных критических параметров процесса (CPPs) с целью оценки влияния обнаруженных факторов на многомерные атрибуты критического качества (CQAs) продукта на стадии первоначального проектирования процесса производства. Метод представляет новый подход к обеспечению качества продукта, который учитывает проблему ко-кластеризации массивов данных CPPs для определения каузальной связи с CQAs. Используется технология неметрического многомерного шкалирования (NMDS) для определения исходных параметров ко-кластеризации. Ключевые слова: качество через дизайн; критические атрибуты качества; ко-кластеризация; многомерный статистический анализ
Context. This paper presents a method for solving the problem of detecting and taking into account the influence of various (external and/or internal) factors on extreme and risky values of the multivariate observed parameters (covariates) of technological and/or diagnostic processes. Taking into account external and internal influence factors on covariates, by analogy with critical process parameters, is a significant addition to the extreme values statistics and the estimations the influence of the variability of process's covariates on the expected losses, i.e. value at risk. Risk-oriented analysis is an actual tool for the data behavior investigation of the multivariate observations of process's parameters. Objective. To disclose a method for detecting and taking into account the factors influence on the distribution functions parameters of the observed extreme values of process's covariates and determine the influence of these distribution functions parameters on estimates of risks values. Method. The method consistently uses: the procedures of multivariate statistical cluster analysis, transformation the matrix of observed extreme values of process's covariates into data frame with factor variables, estimation the extremal index and distribution functions parameters of nonclustered and clustered the observed extreme data of covariates and estimation the risk values on the calculated values of distribution functions parameters. The proposed sequence of actions is aimed at implementing the information technology of statistical causal analysis of the influence of factors on the variability of process's covariates and their risk values due to the application of the clustering procedure for observed multivariate extreme values of covariates. The method is implementing the R-language packages software. Results. Clustering of the multivariate observed extreme values of process's covariates allows to identifying the influence of environmental (manufacturing) factors and estimates the covariates' risky values taking into account of this influence. Conclusions. The method is an information technology of statistical causal analysis of factors influence on the variability of process's covariates and theirs risk values due to the application of the clustering procedure of covariates' multivariate values. The prospect of further research is to improve the methods of causal multivariate statistical analysis of the various factors influence on the exogenous and endogenous parameters of manufacturing and other processes in order to reduce the variability of these parameters and, as a result, minimize the risks.
Many practical tasks of data multivariate statistical analysis from the standpoint of a risk-oriented process approach (in accordance with ISO 9001: 2015, 31000: 2018) requires the definition of the risk values for the dependent exogenous variables of some processes. This paper proposes the method, which consist of original stages sequence for calculating value-at-risk (VaR) or conditional-value-at-risk (CVaR) of dependent exogenous variables, presented of the extreme data frame of critical manufacture process parameters or other parameters, for example, extreme data of environmental monitoring and etc. Risk analysis method by the extreme data of dependent exogenous variables, presented of the data matrix, uses the result of solving the formalized problem of defines the tails parameters of the joint distributions of exogenous variables as components of a bivariate random variable. It can be argued that the tails parameters of the joint distributions of dependent exogenous variables make the validated corrections of the VaR and CVaR estimates for such variables. This method expands the practical application of extreme value theory for the value at risk analysis of any dependent variables as process parameters.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.