Abstract. This article develops a methodological framework for applying predictive learning analytics in AI-based adaptive digital education. The study focuses on early identification of learning difficulties, individualized support, and data-informed teacher decision-making. In digital learning environments, students generate continuous traces of activity, including assessment results, task completion time, error patterns, forum participation, video viewing behavior, and self-regulation indicators. When these data are interpreted pedagogically, they can help prevent academic failure before it becomes visible in final assessment. The article proposes a predictive adaptive methodology consisting of diagnostic data collection, learner risk modeling, intelligent recommendation, differentiated intervention, teacher dashboard interpretation, and continuous profile updating. Mathematical models are presented for calculating a Learner Risk Probability, an Intervention Priority Score, and a Predictive Methodological Effectiveness Index. The research results demonstrate that predictive analytics strengthens adaptive learning by transforming digital education from a reactive assessment model into a preventive pedagogical support system. The proposed framework emphasizes that artificial intelligence should assist, not replace, the teacher: AI detects patterns, while the teacher validates recommendations, organizes meaningful support, and ensures ethical use of learner data.
Publication Date: 2026-06-21