This study explores student engagement and academic performance within a Digital Textbook platform for K-8 students in South Korea. The platform comprises an “Individual Learning Space,” offering digital enhancements to traditional textbook functionalities, and a “Community Space,” promoting interactive forums and collaborative activities. Using a dataset of 19,332 students, with 1,017 actively engaging in both subsystems, the study employs Latent Class Analysis to classify learners into four distinct clusters: Collective Internalization (n=206), Divergent Exploration (n=67), Peripheral Acceptance (n=390), and Productive Collaboration (n=354). The integration of log data with the 21st century Learning Competency survey reveals that the Divergent Exploration cluster shows significant performance gains in problem-solving and ICT literacy, while the Peripheral Acceptance group experiences competency declines. Random forest analysis highlights the pivotal role of community-related features, such as recommending posts, group activities, and replying, in enhancing cognitive, social, and affective performance. These findings emphasize the importance of personalized, AI-driven instructional scaffolding to optimize learning outcomes for diverse student profiles. 

<i>Keywords : Digital Textbook, Learner Profiles, Learning Analytics</i>