Data-Driven Adaptive Curriculum Design Using Clustering and Learning Analytics

The increasing heterogeneity of learners in digital learning environments has exposed fundamental limitations of static, one-size-fits-all curriculum designs. While adaptive learning systems have been widely explored, many existing approaches focus on surface-level personalization and fail to restructure curriculum pathways based on empirical learner behavior. This study proposes a data-driven adaptive curriculum design framework that integrates learning analytics and unsupervised clustering to enable curriculum-level adaptation in online learning systems. Learning interaction data were collected from a learning management system and transformed into multidimensional learning analytics indicators capturing engagement, assessment behavior, completion patterns, and self-regulation. Using clustering techniques, learners were grouped into distinct behavioral profiles without predefined labels. The results reveal stable and interpretable learner clusters characterized as engaged learners with uneven mastery, struggling learners with weak self-regulation, and efficient mastery-oriented learners. These clusters were systematically mapped to differentiated curriculum pathways involving variations in content granularity, pacing strategy, feedback intensity, and assessment difficulty. Empirical evaluation demonstrates that curriculum differentiation leads to divergent but equitable learning outcomes across pathways. Learners in accelerated pathways achieved the highest mean outcome scores with low variance, while learners in scaffolded pathways maintained competitive completion rates despite greater performance variability. Importantly, the adaptive curriculum did not amplify learning disparities; instead, it supported sustained engagement and progression across heterogeneous learner groups. System-level analysis further confirms that the proposed closed-loop adaptive framework remains stable and interpretable, enabling continuous refinement through learning analytics feedback. These findings contribute evidence that clustering-based learner profiling can be operationalized into curriculum engineering decisions, advancing adaptive learning from reactive personalization toward structured, data-driven curriculum design. The proposed framework offers practical implications for scalable, accountable, and pedagogically grounded adaptive learning systems.