Assessing Blended Learning Competencies Using Multimodal Data and Dual Regression

Abstract In the digital era, the educational sector is experiencing a profound transformation driven by the potential of digital technologies to enhance personalized learning, improve educational efficiency, and promote equity. However, the use of multimodal data for competency assessment remains underexplored, particularly in the context of integrating online and offline learning activities. This study proposes a multimodal data-based competency assessment model that integrates online and offline learning activities within a four-tier evaluation framework. Using ridge regression and support vector machine (SVR), this study demonstrates how multimodal data can improve the scientific rigor and effectiveness of competency assessments. Key findings emphasize the importance of rational supervision mechanisms for assessment validity and the role of foundational skills in fostering student innovation. These insights contribute to the intelligentization of educational evaluation and offer practical implications for teaching management.