Development of the mechanism of intellectual management of “student-lecturer” relations in the space of virtual education with the use of neural networks

Research objective . The paper deals with the solution of the problem of regulation of lecturer and student relations using neural networks for bimodal electronic universities. First, the approaches of other authors to this topic are explored. It is known that in bimodal electronic universities, students are trained both in the traditional order and outside the classroom (in distance form). Studies show that for traditional universities there are different approaches with the use of expert systems, genetic algorithms to solve this Development of the mechanism of intellectual management of “student-lecturer” relations  in the space of virtual education with the use of neural networks problem. However, the problem is not solved using neural networks, as well as for students who have an off-campus. The relationship between lecturers and students in universities is regulated by the schedule of lessons. Materials and methods . Schedule of lessons is available in various combinations. The combination of the schedule of lessons is determined once in the beginning of each semester for traditional education, but these combinations are constantly changing in the distance learning environment. These changes are due to the requirement of independence of time and space for distance learning environments. This indicates the relevance of the problem and the compatibility of the use of neural networks to solve the problem. In the presented paper, to solve the problem, the schedule of classes is considered as a matrix. As you know, it is easier to process matrix elements when it is in the form of numbers. To this end, the subjects to be taught in the group are numbered according to the lecturers who will teach them. Then the schedule of lessons is compiled in an arbitrary combination of matrices 3×5 in accordance with the weekly schedule of lessons. Methods of research . The solution process is implemented in the MATLAB environment. To simplify the solution process, the matched matrix is converted to a single row matrix using the RESHAPE command. Then, all combinations of a single row matrix are obtained by applying the PERMS function. As a result, we obtain 15!×15 – dimensional matrix. This matrix is used as the target matrix of the neural network. After that, the weekly schedule of each lecturer’s training is also compiled as a matrix of 3×5, called the “lecturer employment matrix” and is marked as M (I) according to the numbers of the lecturers. Elements of “lecturer employment matrices” can get the value “0” or “İ”. The value of the matrix element is “0”, indicates that the lecturer is busy, and “İ” is free. At the next stage, the “employment matrix” for each lecturer turns into a linear matrix 1×15. Then an input matrix is constructed that combines the elements of the “employment matrix” of lecturers with each element of the target matrix. The weight coefficients for neurons are defined as the difference from the target matrix of the input matrix. Results of the study. Thus, the problem is introduced into a two-dimensional linear equation. Then, the neural network model is selected, tuned and trained in accordance with the conditions of the problem. Conclusion. In the end, the network is tested with input prices that correspond to the interests of the user. The schedule of lessons according to the received results is represented. At the end of the paper, the method side for the distance learning process is useful.