- Пікуляк, М.В. (orcid.org/0000-0003-2192-1899), Кузь, М.В. (orcid.org/0000-0002-9875-1579) and Ворощук, О.Д. (orcid.org/0000-0003-0835-544X) (2022) Іmprovement of information technology of distance education system construction with the use of hybrid learning algorithm Інформаційні технології і засоби навчання, 2 (88). pp. 167-185. ISSN 2076-8184
Text
Микола Васильович Пікуляк.pdf - Published Version Download (1MB) |
Abstract
The theoretical analysis of neuro-fuzzy systems is performed in the article, as well as their main characteristics are generalized and systematized, the peculiarities of known development algorithms are detailed and the relevance of their use for construction of computerized educational programs is substantiated. The structural model of the adaptive educational system is presented and the system of input educational rules modeled on the results of the conducted experiment is described. In order to determine the assessment of the current level of student learning, a number of qualitative indicators (depth of study, degree and quality of learning) were introduced, the use of which allowed to ensure the completeness of the base of input rules for fuzzy inference. A method based on a fuzzy neural network for constructing an adaptive module of a remote knowledge transfer system is proposed, the application of which makes it possible to increase the speed and accuracy of calculations at the stage of determining the training mode according to the current level of student knowledge. An adaptive mechanism for constructing an individual learning trajectory in the distance education system based on the fuzzy Mamdani neural network has been implemented. A hybrid algorithm for learning a neural fuzzy network has been developed and the stages of its operation are given. Peculiarities of application of hybrid algorithm for determination of educational mode are investigated and advantages of its use by parallel and simultaneous specification of network parameters are established. A block diagram of a hybrid algorithm of an adaptive training module is proposed, which allows to modify the output rules in the process of network learning according to the given learning accuracy. An experimental study of the application of a hybrid algorithm using a fuzzy neural network ANFIS in the program MATLAB was conducted and its confirmed the effectiveness of the proposed technology. Prospects for the use of mathematical tools of neural network technologies in the study of adaptive characteristics of automated learning systems are determined.
Downloads
Downloads per month over past year
Actions (login required)
View Item |