PREDICTION OF USERS ACTIVITY IN THE MOODLE PLATFORM BASED ON MACHINE LEARNING METHODS

Abstract

The article presents the creation of a machine learning model for activity prediction based on the Scikit-learn library. The model allows to predict activity based on data about the actions of users of the Moodle platform. The program was developed in the Python language in the PyCharm software development environment. The amount of data taken for processing was 1000 samples of users from the Moodle database. Classification was used as the machine learning task, and the random forest method was used as the method. Random forest copes well with overfitting problems and scales well for large data sets. It is also an ensemble method that combines several decision trees to achieve better accuracy and stability compared to single decision trees. It has been verified that the random forest method has relatively high accuracy and low duration of the learning process. The overall accuracy of the developed model was calculated, which is 83%. Increasing the accuracy of the obtained model is possible due to the expansion of the source data, which requires the creation of appropriate applications (plugins) for the Moodle platform. It has been verified that the use of the Random Forest machine learning method for classification tasks is well suited for predicting the category or class of a new sample of user activity based on its characteristics.

The presented information shows that choosing the Scikit-Learn library will allow to create an effective model of data processing and prediction of results. The statement about the feasibility of choosing the Scikit-Learn library also coincides with the result of the analysis of modern libraries used for the development and training of machine learning models. The use of the created model for forecasting will allow to quickly analyze the activity of users and form, if necessary, appropriate ratings.