MKG-Driven Student-Resource Matching & Adaptive Intervention System: A Framework for Reducing Conceptual Errors and Enhancing Engagement via Multimodal Learning Analytics

Abstract

This study proposes MKG-Driven Student-Resource Matching & Adaptive Intervention, a framework designed to reduce conceptual errors and enhance engagement through multimodal learning analytics. The framework integrates two core components: Learning Analytics & Tracking, which employs an LSTM-based knowledge tracing model to predict mastery probabilities and computes a Learning Engagement Index from multimodal behavior data, and Intervention Mechanism, which dynamically triggers concept-specific micro-video recommendations when group error rates exceed predefined thresholds. The system generates personalized learning pathways for individual gaps while providing class-level heatmaps to highlight knowledge gap hotspots, thereby enabling targeted instructional adjustments. A Multimodal Learning Analytics Dashboard synthesizes these insights through visualizations such as radar charts, donut charts, and heatmaps, offering educators real-time monitoring capabilities. The implementation workflow involves data fusion, gap detection, resource matching, and teacher support, ensuring seamless integration of heterogeneous learner data. Our approach addresses the challenge of scalable, data-driven interventions by combining fine-grained knowledge tracing with adaptive resource allocation. Experimental results demonstrate its effectiveness in reducing error rates and improving engagement, with the framework's novelty lying in its dual focus on individual and group-level analytics. The system's significance extends to practical educational settings, where it bridges the gap between theoretical learning models and actionable pedagogical strategies.