ASSESSMENT OF THE EFFICIENCY OF THE SUCCESS PREDICTION MODEL USING MACHINE LEARNING METHODS

Abstract

The work evaluated the created model for predicting the success of users of the Moodle educational platform in order to determine the effectiveness and expediency of its use. The determination of the efficiency and quality of the model was performed according to the following indicators: sensitivity, specificity and balanced accuracy. Also, a ROC curve was constructed to reflect the classifier's ability to correctly recognize positive classes and reject negative classes when the threshold value changes, and the AUC (Area Under Curve) was determined. Methods of assessing success risks and requirements for creating models based on machine learning methods are analyzed. On the basis of user data from the database of the Moodle educational platform, indicators affecting the success of students were formed. A logistic regression model was built for predicting success, which was then tested in practice. The created model makes it possible to predict the success of students with an accuracy of 84%. The overall efficiency of the model is 89%. It was established that the Scikit-learn library provides an opportunity to create an effective model for solving classification problems in machine learning. The use of a logistic regression model to classify the success of users of the Moodle platform will allow creating a model that allows you to predict the success of students based on the collected data. Using machine learning methods and Python libraries, the quality and efficiency of the model was determined. A solution to the problem of predicting students' success is proposed by creating a model based on machine learning methods and the Moodle platform. A Python program was developed to analyze the data of users of the Moodle platform.

The presented information shows that choosing the Scikit-Learn library will allow creating an effective model for processing data and predicting results. The use of the created model for forecasting will allow to quickly analyze the success of users and form, if necessary, appropriate ratings. The forecast obtained with the help of the model will be useful both for teachers and educational institutions. Because it will allow planning changes in educational programs and materials, as well as the educational process as a whole.