CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
HYBRID INTELLIGENT TECHNOLOGY OF DATA CLASSIFICATION
 |
Read | Demidova Liliya A., Egin Maksim M. HYBRID INTELLIGENT TECHNOLOGY OF DATA CLASSIFICATION // Caspian journal : management and high technologies. — 2018. — №1. — pp. 56-68. |
/56-68.jpg)
Demidova Liliya A. - Doc.Sci. (Engineering), Professor, Ryazan State Radio Engineering University, 59/1 Gagarin St., Ryazan, 390005, Russian Federation, liliya.demidova@rambler.ru
Egin Maksim M. - Student, Ryazan State Radio Engineering University, 59/1 Gagarin St., Ryazan, 390005, Russian Federation, eginmm@gmail.com
The paper considers the problem of binary classification of the data sets of various nature, presented by numerical values of characteristics. This problem is solved with use of data mining tools. The aim of the paper is to create the hybrid intellectual technology for data classification (HITDC) based on the joint use of the SVM algorithm and the Parzen windows. The classifier, based on the Parzen windows (CbPWM), improves the accuracy of data classification performed using the classifier based on the SVM algorithm (CbSVM). The CbPWM applies to data that can be both correctly and erroneously classified using the CbSVM. These data are located in the experimentally defined subareas near the hyperplane separating the classes. The technology implies the use of default parameters values for CbSVM, while the suboptimal parameters values of the CbPWM are determined using the genetic algorithm. The paper presents the results of experimental studies, confirming the effectiveness of the proposed hybrid intellectual technology for data classification.
Key words: классификатор на основе SVM-алгоритма, опорный вектор, тип функции ядра, параметр функции ядра, параметр регуляризации, классификатор на основе метода окна Парзена, метрика расстояний, гибридная технология, генетический алгоритм, вычислительные эксперимен