A distinguishing feature of intelligent game-based learning environment is its capacity for assisting stealth assessment. Stealth assessment gathers data regarding student competency in an invisible way and enables drawing valid inferences with respect to student knowledge. Stealth assessment might radically extend the impact and scope of learning analytics. Stealth assessment describes the unobtrusive assessment of learner by using emergent data from the digital traces in electronic learning environment via machine learning technology. This study presents a new stealth assessment model using ensemble learning for inferring the student competency in a game-based learning (GBL) environments, named ELSAM-GBL technique. To perform automated and accurate stealth assessment, this study focuses on the design of ensemble learning model by the incorporation of three DL models, namely gated recurrent unit (GRU), sparse auto encoder (SAE), and vanilla recurrent neural network (RNN). At the same time, the hyperparameter tuning of the DL models takes place using the atomic orbital search (AOS) optimization algorithm, which helps in improving the ensemble learning process. To demonstrate the enhanced stealth assessment performance of the ELSAM-GBL technique, a comprehensive experimental analysis is conducted. The comparative study shows the enhanced performance of the presented ELSAM-GBL technique over other DL models in terms of different metrics.

智能游戏学习环境的一个显着特点是其辅助隐形评估的能力。隐形评估以无形的方式收集有关学生能力的数据，并能够对学生的知识得出有效的推论。隐形评估可能会从根本上扩展学习分析的影响和范围。隐形评估是指通过机器学习技术，利用电子学习环境中数字痕迹中的新兴数据，对学习者进行不显眼的评估。本研究提出了一种新的隐形评估模型，使用集成学习来推断学生在基于游戏的学习 (GBL) 环境中的能力，称为 ELSAM-GBL 技术。为了执行自动化和准确的隐形评估，本研究重点关注结合三种深度学习模型的集成学习模型的设计，即门控循环单元（GRU）、稀疏自动编码器（SAE）和普通循环神经网络（RNN）。同时，使用原子轨道搜索（AOS）优化算法对深度学习模型进行超参数调整，这有助于改进集成学习过程。为了证明 ELSAM-GBL 技术增强的隐身评估性能，进行了全面的实验分析。比较研究表明，在不同指标方面，所提出的 ELSAM-GBL 技术比其他深度学习模型具有增强的性能。