Application of Learning Analytics technologies to predict students' educational outcomes in computer science

Abstract

In the context of the digital transformation of higher education, learning analytics technologies are a result of the personalization of learning and its improvement. In this context, the study of computer science is becoming increasingly relevant, where the development of algorithmic thinking and practical programming skills requires continuous and detailed monitoring of educational activities. The aim of the study is to develop and validate a model for predicting student educational outcomes based on learning analytics data within information science disciplines. The study utilizes educational data mining methods, machine learning algorithms (classification and regression), and statistical analysis of student interaction logs with learning management systems (LMS) and interactive programming environments. The study identified the most significant predictors of academic performance, including regular access to materials. The time spent completing practical assignments is recorded separately. The number of code compilation attempts and patterns of reference resource access form the basis for analyzing educational activity. The developed predictive model identifies at- risk students early. Accuracy in the early stages of the semester exceeds 85%. The integration of predictive analytics into the computer science curriculum shifts the focus from retrospective assessment to proactive pedagogical intervention. Individual educational trajectories are subject to timely adjustments. The study's results are applicable to the design of adaptive educational environments. They remain applicable to the development of pedagogical decision support systems and the improvement of computer science teaching methods at universities.