Intellectual prediction of student performance: opportunities and results

摘要

Intellectual tools for analysis and forecasting are widely used in various fields - economics, medicine, technology, and linguistics. This article examines the possibilities of neural network forecasting of student performance. A general statement of the research object is formulated. Comprehensively considered and classified factors affecting student performance are pre-university, university and psychophysiological ones. The features of the collection of information in Russian universities for intellectual analysis and forecasting are considered. Using the example of the database of the Financial University under the Government of the Russian Federation and additional information obtained by survey about factors, not present in the database, significant factors were determined using the correlation analysis toolset of the IBM SPSS Statistics statistical analysis package, which made it possible to reduce their number almost fourfold and record the regression model in a simpler form. Further research was carried out using a Deductor Studio analytical platform for intellectual processing and knowledge extraction. A multilayer neural network with nine entrance signs and one or two effective ones was built and trained. The effective entrance signs were taken as the results of the first year students taking senior exams. The research results showed that the predicted values of progress do not differ significantly from the actual ones. Consequently, neural network machine study technologies provide intelligent prediction of progress based on an analysis of the preceding factor signs - of both the first and subsequent years. The directions of further research with the use of modern means of machine study are outlined.