Edge–Cloud Adaptive Ensemble Framework (ECAEF) for Real-Time Student Performance Prediction

Abstract

Effective learning analytics are essential to early and accurate prediction of student performance, which can be used to intervene in time and make informed decisions. This paper deals with the shortcomings of current machine learning models, which include offline batch learning and high inference latency, with the presentation of the Edge-Cloud Adaptive Ensemble Framework (ECAEF). The framework enables prediction of real-time performance in streaming data environments by combining edge intelligent with cloud coordination. It consists of four modules: the Edge Learning Stream Collector (ELSC) to maintain data acquisition continuously, the Incremental Dual-Model Learner (IDML) that integrates a Customized Random Forest (CRF) and a Legacy Recurrent Neural Network (LRNN) to maintain stable learning, the Adaptive Fusion Controller (AFC) to dynamically regulate itself based on performance metrics, and the Edge-Cloud Synchronizer (ECS) to exchange knowledge between nodes.