期刊名称：Eastern-European Journal of Enterprise Technologies
印刷版ISSN：1729-3774
电子版ISSN：1729-4061
出版年度：2016
卷号：4
期号：4
页码：20-28
DOI：10.15587/1729-4061.2016.75679
语种：English
出版社：PC Technology Center
摘要：Modeling natural systems and processes often requires making decisions under conditions of data uncertainty. There are a number of applications, in which obtaining data about the system is a time- and resource-consuming process. That is why it is advisable to develop decision making mathematical models and methods capable of functioning on input data that has gaps.In the scope of this research an aggregated interval type-2 fuzzy data classification model was built. Interval type-2 fuzzy set mathematics operates membership grades in an interval form. This allows to take uncertainties intrinsic for an input vector into account, and enables working with an input vector that has gaps instead of some of the values. The proposed system supports involving one or more experts on early stages of the decision making process, more specifically in the phase of informative feature extraction. A feature set provided by every expert generates a separate model with an interval output. Such an approach achieves incorporating expert’s knowledge and experience into the decision making process along with information accumulated in the experimental data set. The results of multiple models are subsequently aggregated into a single interval, which is a generalized interval estimation of the system’s status based on data available at the moment, and allows to get an idea of the uncertainty associated with the decision taken.It is also allowed to integrate third party models based on other decision making methods and technologies, such as a modification of the PCM cluster analysis method with interval membership grades.The aggregated model was adapted for approximate evaluation of groundwater extraction perspective on different states of hydrogeological exploration. It was determined that in some cases it can save significant cost and human resources.
