DIGITAL LIBRARY
SUCCESS PREDICTING FOR STUDENTS OF ONLINE COURSES USING NEURAL NETWORKS
1 Ural Federal University (RUSSIAN FEDERATION)
2 Forschungszentrum J├╝lich GmbH (GERMANY)
About this paper:
Appears in: INTED2020 Proceedings
Publication year: 2020
Pages: 6547-6554
ISBN: 978-84-09-17939-8
ISSN: 2340-1079
doi: 10.21125/inted.2020.1746
Conference name: 14th International Technology, Education and Development Conference
Dates: 2-4 March, 2020
Location: Valencia, Spain
Abstract:
Online education develops dynamically in Russia and in the world. However, its use is associated with a number of specific problems caused by a large number of students and the lack of personal contact with them. Among these problems are - loss of motivation and early termination of education, difficulties in mastering certain significant elements of the course and, as a result, low learning outcomes of the course as a whole. Analysis of the progress in training and forecasting the results of the course are important tasks in solving the above problems.

The paper considers a model for analyzing current performance and predicting the results of final tests using elements of artificial intelligence. A data model based on the digital footprint of online courses students has been built. An algorithm for analyzing and predicting the effectiveness and success of training based on neural networks such as Kohonen SOM (self-organizing map) and multilayer perceptron has been developed. Based on the methods of information theory, the mutual information contained in the data on current performance and the results of the final testing were evaluated. A model has been developed for estimating the theoretical limits of forecasting accuracy for the results of final testing based on learning analytics.

The models and algorithms are tested on real data from the online course of the Ural Federal University, presented on the National Open Education Platform. The results of the study can be used to design support systems for students in online courses, as well as to improve the methods of online learning and pedagogical design.
Keywords:
Online education, data model, digital footprint, forecasting, artificial intelligence, neural networks.