Research and forecasting of educational process parameters by using artificial intelligence

Abstract

In this paper we present the results of an interdisciplinary research based on the application of big data, date science, artificial intelligence and machine learning methods in educational analytics. Artificial intelligence techniques applied for the analysis of depersonalized data stored in the information and analytical system "E-education in the Republic of Tatarstan" from 2015 to 2020. BigData technologies were used in this work to perform high-performance computing related to initial preprocessing of raw data in computation cluster. By using the methods of artificial intelligence, we modelled one of the most important stages in the formation of the educational trajectories of schoolchildren, associated with the fact that after the 9th grade, schoolchildren either continue their studies in high school (grades 10-11), or move to the professional educational organizations. As the input data for neural network training, we used a vector containing the average marks for all quarters of pupils, obtained by using high-performance Dask-based cluster data processing system from initial raw data. We concluded that multi-layer neural network with two hidden layers was able to predict the pupil’s pass to 10th grade, and achieved the best performance with classification accuracy exceeding 70%. Also, the performance of trained neural network had been analyzed by visualization of Receiver Operator Characteristic (ROC)-curve and by calculation of recall, precision, specificity and area covered by the ROC-curve (AUX) parameters.