Mining of Online Learning Outcomes: A Stacking Heterogeneous Ensemble Learning Approach

The rapid development of information technology has transformed online learning platforms into pivotal channels for individuals seeking to acquire new knowledge and skills. As these platforms accumulate vast amounts of user behavioral data, they offer a unique opportunity for predicting user online learning behaviors. However, existing prediction models predominantly rely on singular algorithms, indicating substantial room for enhancing prediction efficacy. In this study, we propose a novel heterogeneous ensemble framework that leverages stacking strategy, aiming to forecast user behaviors on online learning platforms. Through model evaluations conducted on a popular online learning platform in China, our findings demonstrate that the proposed heterogeneous ensemble framework outperforms both singular algorithmic models and homogeneous ensemble approaches. Our approach holds promise for aiding platform operators in forecasting user learning outcomes, thereby facilitating the design of tailored incentive strategies to enhance user engagement and motivation.