Educational data mining focuses on exploring increasingly large-scale data from educational settings, such as Learning Management Systems (LMS), and developing computational methods to understand students' behaviors and learning settings better. There has been a multitude of research dedicated to studying the student learning process, leading to multiple commonly cited frameworks and theories that characterize students' learning behaviors. However, the most recent focus of interest argued that developing efficient models with actionable suggestions that help understand student learning behaviors and contribute to their academic outcomes is needed. Existing studies investigated various individual, emotional, and social factors related to student learning behaviors by analyzing fine-grained data samples collected by LMS at the student level but fail to incorporate the dynamics of learning behaviors and fail to examine the whole image of the student learning lives. For instance, most research in student learning behavior focuses on specific courses and related academic performance but did not consider that students always take multiple courses simultaneously. Therefore, the implications and knowledge from the static understanding of students' learning behavior in an isolated specific course may be limited to generate actionable strategies to help students. This dissertation was motivated to explore large-scale data, especially for examining the learning behavior's dynamics and developing student-centric models. The resulting knowledge has been recognized by publications in top-level international conferences in educational data mining and artificial intelligence in education. The first research attempt of my research introduced a new computational method based on psychological theories in affect dynamics to track dynamic student behaviors and developed a novel explanation method based on Local Interpretable Model-Agnostic Explanations (LIME). The second research attempt focused on developing computational models at the student level to predict their academic performance based on LMS data at the University of Maryland, Baltimore County. The findings from this work showed that student login volume and their prior performance significantly impact student performance. Additionally, this research focused on exploring causal relationships between student LMS behaviors and their academic performance. The causal analysis strengthened our findings in computational modeling by showing a significant cause-and-effect relationship between student login behaviors and their academic performance. The conclusions from this work will empower intervention techniques that improve student emotion regulation capabilities. The student-centric models developed in this study reported the positive impact of student login behaviors on their academic performance. This understanding will enable LMS developers and school administrators to design and develop interactive systems that deliver course content effectively. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com.bibliotheek.ehb.be/en-US/products/dissertations/individuals.shtml.]