Abstract
Blended learning has become the mainstream model in higher education. However, existing studies on learners' behavioral characteristics mostly rely on short-term and small sample data, and lack in-depth exploration of the intrinsic relationship between objective behavioral logs and subjective self-assessments. This study, based on ten-year longitudinal tracking data (N=3693), integrates platform objective learning logs and learners' subjective self-assessment scales, uses stepwise regression to screen key behavioral variables, and then employs K-means clustering to identify learners' behavioral types. The results show that there is a systematic separation between objective academic performance and subjective self-assessment results, and gender and academic background are important moderating variables; the impact of learning behaviors on learning outcomes exhibits a non-linear characteristic, and interaction quality and self-regulation ability are the core levers beyond "time accumulation"; self-assessed learning behaviors are generally high and have relatively small individual differences, which contrasts sharply with the significant individual differences in objective behavioral indicators. This study provides a methodological example for multi-source and longitudinal learning analysis and has important practical implications for the design of adaptive learning systems and the formulation of precise teaching intervention strategies.