Real Time Digital Twin Framework for Big Data-Driven Online Education Ecosystem

INTRODUCTION: The rapid expansion of online education platforms has created unprecedented opportunities for personalized learning but also presents challenges in monitoring student engagement, learning outcomes, and instructional quality in real-time. OBJECTIVES: This research proposes a real-time digital twin framework for a big data-driven online education ecosystem, designed to continuously capture and analyze large-scale educational data that enhances learning and teaching effectiveness. METHODS: The digital twin online learning dataset with 345 students is obtained. The obtained data are preprocessed by the data cleaning for removing duplicate entries and z-score normalization to normalize the numerical features in the dataset. By integrating big data analytics and Deep Learning (DL) techniques based on an Adaptive Monarch Butterfly Optimized Graph convolutional with Long Short-Term Memory (AMB-GC-LSTM) that combines Long Short-Term Memory (LSTM) networks for sequential student behavior prediction and Graph Convolutional Networks (GCNs) for modeling collaborative learning relationships, the system enables accurate prediction of student engagement, performance trends, and early identification of at-risk learners. RESULTS: The AMB is employed to optimize the GC-LSTM parameters for higher accuracy and cost computation reduction. Experimental evaluation demonstrates that the digital twin-driven ecosystem improves engagement, learning outcomes, and teaching efficiency. CONCLUSION: Comparison results provide an enhanced accuracy (0.965), precision (0.945), recall (0.974), and F1-score (0.959) with the Python implementation. This research presents a novel approach for integrating digital twin technology, big data analytics, and deep learning to create an intelligent, responsive, and scalable online education ecosystem capable of supporting continuous improvement in teaching and learning.