DESIGN AND EVALUATION OF REAL-TIME ADAPTIVE LEARNING ALGORITHMS FOR PERSONALIZED K-12 CURRICULUM OPTIMIZATION USING STUDENT PERFORMANCE ANALYTICS

Maduabuchukwu Augustine Onwuzurike*, Joy Onma Enyejo, Amina Catherine Peter-Anyebe

ABSTRACT

The growing heterogeneity of learner abilities and pacing in K–12 education has exposed fundamental limitations in static and batch-oriented curriculum delivery models. This study presents the design, implementation, and evaluation of a Real-Time Adaptive Learning (RT-AL) framework for personalized K–12 curriculum optimization using continuous student performance analytics. The proposed system integrates event-driven data ingestion, online learner state estimation, sequence-based predictive modeling, and constrained curriculum policy optimization to enable instructional decisions within live learning sessions. Learner mastery was modeled as a continuously updated probabilistic state derived from correctness, response latency, and contextual interaction signals, allowing dynamic adjustment of content sequencing, difficulty, and pacing. Empirical evaluation was conducted against four baseline approaches static curriculum sequencing, batch adaptive learning, rule-based personalization, and random assignment using predictive accuracy, area under the ROC curve (AUC), normalized learning gain (NLG), time-to-mastery reduction, engagement retention, and system latency as core metrics. Results show that RT-AL achieved superior predictive performance (accuracy = 0.91, AUC = 0.94), substantially higher learning gains (NLG = 0.42), and significantly lower response latency (120 ms) compared to all baselines. Personalization effectiveness analysis across diverse learner profiles revealed particularly strong gains for low prior knowledge and at-risk learners, while maintaining high calibration and enrichment alignment for high achievers. Curriculum-level analysis demonstrated that real-time adaptation enabled greater flexibility, deeper personalization, improved assessment alignment, and sustained learner engagement relative to static designs. The findings establish that real-time, analytics-driven adaptation is not merely an optimization enhancement but a structural requirement for effective personalized learning at scale. This study contributes a validated technical architecture, learner modeling approach, and evaluation framework that collectively advance the deployment of real-time adaptive intelligence in K–12 education.

Keywords: Adaptive Learning Algorithms; Personalized Curriculum; Student Performance Analytics; Real-Time Learning Systems; K-12 Education Optimization.