Visual Classification of Data Sets with the Assistance of Experts in the Problems of Intelligent Agents Learning for Incompletely Automated Control Systems
Alena Alexandrovna Zakharova, D.A. Korostelyov · 2019
The paper proposes a method of data classification, which allows replacing nodes, with which operators or experts interact, with their intelligent software analogues in incompletely automated control systems. The classification is based on the algorithm of interactive interaction between an expert and the system, in which the offered visual images of the corresponding vectors from the data set and the reference ones are compared. Using various methods of visualizing data vectors and stricter selection requirements, a method is described that allows to classify all the vectors from the presented data set. The experiment on the visual classification of data on the example of diagnosing an ultrasonic flow meter is also described. The results obtained are compared with other classification methods.