AI-driven predictive models for optimizing mathematics education technology: enhancing decision-making through educational data mining and meta-analysis

This paper explores the challenge of achieving consistent effectiveness in integrating Mathematics Education Technology (MET) in K-12 classrooms, focusing on factors such as technology type, timing, and instructional strategies. It highlights the difficulties novice teachers face in optimizing MET compared to experienced educators, emphasizing the need to better understand the ideal duration and application of MET in various teaching settings. This study proposes using Artificial Intelligence (AI) to predict and optimize MET effectiveness, aiming to enhance student achievement. However, a key challenge is the lack of comprehensive MET databases, prompting the exploration of novel data collection methods and meta-analysis for educational data mining. An AI-based predictive model is developed for MET, analyzing 423 publications on its effectiveness in Chinese K-12 mathematics education. Nine AI-driven predictive models were created, with the best-performing predictive model being eXtreme Gradient Boosting, enhanced with L2 Regularization, Synthetic Minority Over-sampling Technique–augmented Regression (SMOTER), and Active Learning. The proposed model was further refined using Particle Swarm Optimization for hyperparameter tuning and analyzed with Shapley Additive Explanations (SHAP) values to assess feature importance. Numerical results indicated that the duration of MET usage is a critical factor for optimization. A controlled experiment in a Mainland China middle school validated the model’s efficacy, showing that model-guided MET significantly outperformed traditional methods. These findings offer valuable insights for bridging gaps between novice and experienced teachers, promoting educational equity, and providing practical recommendations for improving MET integration in Mathematics education.

Text 123319.pdf - Published Version Available under License Creative Commons Attribution. Download (3MB) Official URL or Download Paper: https://link.springer.com/article/10.1186/s40561-0...

Actions (login required)

View Item