Improving online learning system by means of artificial intelligence

Abstract

Introduction. Online learning is applied in the educational system more and more actively every year. The bulk of learning activities in many programmes and courses is shifting from the classroom towards individual work. The interaction between the student and the teacher is decreasing, which negatively affects the quality of education. The aim of the research is the development of methods and algorithms for assessing the psycho-emotional state of learners and their involvement in the teaching process in online education, as well as evaluation of their impact on the efficiency of training. Materials and methods. More than 100 students of Russian State Agrarian University – Moscow Timiryazev Agricultural Academy and Financial University took part in the survey. The following characteristics were measured during the online teaching: involvement in the educational process, students’ psycho-emotional state. In order to build the training algorithms assessing students’ engagement in online learning, the modernised dataset DAiSEE was used; for assessing the psycho-emotional state – the dataset fer2013. Convolutional neural networks were used as model algorithms. The ROC curve and accuracy parameters were used as metrics for model training quality. The Farrar-Glauber test was used for the analysis of multicollinearity of parameters. To prove the efficiency of the developed methods, the statistical criteria – Fisher’s F-test and Student’s t-test – were used. Results and discussion. The developed models demonstrated an excellent quality of training. The accuracy of recognising student engagement in the learning process exceeded 90%, the accuracy of identifying the emotions in online learning was over 90%. The inculcation of the said algorithms in the online learning system showed due efficiency. It was proved that the control group (23 students) and the experimental group (25 students) differed significantly in statistical terms (temp = 2.53; p < 0.05). The learners’ involvement, their emotions and the actual knowledge of the subject area did not show any strong correlation with each other (FGemp=3.61 < FGcrit=7.81). Conclusion. It was proved that the introduction of artificial intelligence systems in the learning process made it possible to adjust the online learning system with regard for the obtained recommendations on the student’s psycho-emotional state and his/her involvement in the training process. Thus, the interaction between the student and the teacher, although weakened in comparison with face-to-face training, nevertheless remains at a sufficiently high level and allows influencing the students’ motivation and mood, which leads to improved quality of education.