Digital learning tools and learning analytics as resources for the personalization of the educational process and pedagogical decision-making

Abstract

The article is devoted to the theoretical and empirical substantiation of how digital tools and learning analytics can serve as a resource for the personalization of the educational process and support evidence-based pedagogical decision-making in higher education. The aim of the study is to identify statistically significant correlations between parameters of students’ digital footprints and their academic performance, to develop behavioral profiles of learners, and to test the effectiveness of a system of personalized recommendations. The empirical base was formed over a three-year longitudinal experiment (2020-2023) at a large technical university involving 1,748 undergraduate students; data were collected in an integrated digital environment based on the LMS Moodle platform with webinar, interactive simulator, and proctoring modules, with more than 5 TB of logs processed in total. Methods of descriptive statistics, correlation and regression analysis, clustering (k-means) to identify typical behavior patterns, as well as quasi-experimental comparison of groups using various digital tools, were employed. It was shown that the decisive factor for success is the nature of activity: the number of completed interactive tasks demonstrates a strong positive correlation with final academic performance (r=0.723; p 0.001), whereas passive content consumption (viewing video lectures, formal participation in forums) has significantly lower predictive power. The introduction of interactive simulators and adaptive testing provides an increase in final grades of up to 20% compared with the traditional model of online courses. Cluster analysis of the digital footprint made it possible to identify four stable student profiles ( systematic and active , minimalist performers , procrastinators , passive observers ) with significantly different academic performance and levels of academic risk. Based on these profiles, a system of personalized recommendations was developed and tested, the implementation of which led to an average performance increase of 10,89%, a 27,81% reduction in the proportion of at-risk students, an increase in access to supplementary materials by more than 1,5 times, and a decrease in the average task completion time. Discussion of the results allows digital tools and learning analytics to be characterized as a foundation for the transition from a mass unified model to truly student-centered, adaptive education and outlines prospects for the use of more advanced AI algorithms and user-friendly dashboards for instructors in future intelligent educational ecosystems.