Abstract:

Today's educational system must keep pace with technological development, be fast and efficient. The percentage of students who fail a course is perhaps the best measure of the efficiency of an educational system, and it should certainly be kept as low as possible. It can be assumed that some students will be able to pass the course with the help of individualised educational interventions, but the question is how to philtre out this at-risk group before course failure becomes evident. In this paper, we address this question and develop a model for selecting students who are at significant risk of failing the course.

Supervised machine learning methods, also known as predictive methods, predict the values of output variables based on input data. The model is developed using training data in which the values of the input and output variables are defined. The model generalises the relationship between input and output variables and uses it to predict other data sets where only the input data is known. If the output variable is a continuous value, regression algorithms are used.

In this paper, we present the use of an artificial neural network for predicting study dropouts. Data on learner activities from the university e-course Mathematics 2 in STEM were used, e.g., click-based data describing whether, when, and how often learners access resources that provide different views of the content, as well as data reflecting learners' activities in the course. In addition to these data, survey data were used to show learner success in mathematics during high school education. The results obtained will be used for early prediction of student failure, i.e., to identify students who are at risk of not completing the course.

Keywords:

Artificial Neural Network, early prediction, student failure.