In this study, we propose a fundamental model for social learning in order to effectively use micro contents as educational resources, based on personal data analytics and individual modeling. We improve data-driven individual modeling, proposed in our previous study, by introducing the concept of dividual and divide-and-conquer algorithm. We further propose a temporal social network analysis approach for dynamic community mining and tracking, and develop overlap community detection algorithm by spectral clustering based on node convergence degree that takes both the network structure and node attributes into account. We use topic evolution model and temporal social network analysis for topic tracking and assessing and pattern extraction as well. Experiments have been conducted based on different data sets, and comparison with related works and result analysis have shown the effectiveness of our proposed model and approach.