Exploring the synergy of improved convolutional neural networks and attention mechanisms for potential STEM knowledge concept recommendation in MOOCs

Abstract

The multi course association of STEM poses an important challenge to the learning background of learners. Once learners do not have sufficient understanding of knowledge association or do not implement the topological order of knowledge advancement, they are prone to burnout in the learning process, forming serious negative emotions, which is not conducive to learning effectiveness, and even premature dropout. This is clearly a psychological teaching problem, that is our research objectives. This study focuses on the STEM learning behaviors in MOOCs, and explores the deep learning routing. We design one novel method to process the context features and content features for knowledge concept recommendation. Multiple entities, features, and courses enable the construction and optimization of knowledge concept relationships. Then, an attention mechanism is used to achieve the knowledge concept propagation between different entities. The extensive experiments have proved this method might accurately capture potential interests of knowledge concepts, achieve the effective deep learning routing, and explore and guide the positive learning state, reduce or avoid the negative psychological outcomes, such as dropout or low pass rate. The entire study aims to enhance learning outcomes, improve learning motivation, optimize learning behaviors, and provide more effective suggestions for STEM education, that is very important for the interdisciplinary learning in higher education. The whole research might provide key support for tracking possible psychological changes in learners, improving learning behavior trends, and enhance learning quality during STEM learning, fully improve and optimize the learning state, construct effective decisions for positive learning interests.