Investigation of the Accuracy of Machine Learning Methods in Prediction of Students Success

One of the main factors affecting the future opportunities and career development of students is the results of the educational process. A high level of success in education makes it possible to obtain qualified specialists in the future and ensure economic growth. Prediction of success is an urgent task, because a quick response to possible problems allows to increase the student's chances of successfully passing the session, as well as to identify deficiencies that can be corrected. In the work, a study of the accuracy of predicting the success of students based on data on their attendance and grades was conducted. The following machine learning algorithms were studied: Logistic Regression, Support Vector Method, Random Forest. The parameters characterizing the accuracy of the classification are determined, namely: sensitivity, specificity and balanced accuracy. Based on the obtained results, receiver operating characteristic graphs were constructed to assess the ability of the classifier to correctly recognize positive classes and reject negative classes when the threshold value changes. In order to perform forecasting, data on user ratings and attendance for the first and second semesters of the 2021–2022 academic year were exported from the Moodle electronic learning management system database. The total number of student grade and attendance records exported to files was 1,308,262. The creation of predictive models was performed in the PyCharm development environment in Python using the Scikit-learn library. After performing the calculations, it is determined that the random forest method performs the best prediction of success based on the available input data and has a higher accuracy. The obtained accuracy value of 80% and AUC of 73% indicates the quality of the classification model and good discriminating power of the model. The graph of the receiver operating characteristic for the random forest method shows that it has a clearly defined area under the curve that is more curved up and to the left, which confirms the effectiveness of the model. The obtained results can be used as a basis for further research.