Educational Data Mining (EDM) is an emerging field that aims to better understand students' behavior patterns and learning environments by employing statistical and machine learning methods to analyze large repositories of educational data. Analysis of variable data in the early stages of a course might be used to develop a comprehensive prediction model to prevent students' failure or dropout. This will assist instructors in intervening effectively. In this defense report, three major contributions in the field of educational data mining and learning analytics are presented. The first contribution is a framework to statistically compare different instructional modes and investigate their effects on students' academic performance. Our analysis is unique in the literature and is based on the "Rank Percentage" rather than students' course grades. The rank percentage allows us to make the academic scores more comparable. Previous studies have mainly focused on students' course grades to evaluate academic performance. A sample data set of around 500 students attending a Financial Engineering Course (IE201) is used to test the proposed approach. A total of four cohorts of students are considered, depending on the teaching modalities they attended: the traditional face-to-face (F2F) classroom, the transitional semester interrupted by the COVID-19 pandemic, and the two consecutive online semesters (asynchronous and synchronous). We aim to investigate the effect of the transition from F2F classes to online modes on different subgroups of students. Accordingly, we divide each student cohort into subgroups using criteria such as cumulative grade point average (GPA). We then utilize the Mann-Whitney U test for pair-wise comparisons between cohorts in each cumulative GPA subgroup. Our findings highlight that the differences between specific subgroups of students are significant. As a result of the transition to online learning, students with cumulative GPAs greater than 2.90 have been negatively affected. It is, however, hard to draw a firm conclusion for students with cumulative GPAs below 2.90. Our results encourage researchers to investigate the effectiveness of various teaching modes on subgroups of students when comparing populations with different instruction modes. The second contribution of this dissertation is a framework by which students' learning style features can be utilized to enhance their academic performance and detect at-risk students early in the course. A publicly available dataset called Open University Learning Analytics Dataset (OULAD) is used to test the proposed approach. We employ the Felder-Silverman Learning Style Model (FSLSM) as the basis for mapping learners' interaction with the Virtual Learning Environment (VLE) to learning style (LS) features. LSs extracted from the course's first and second quarters and the learners' demographic features are used to train different prediction models to classify the second quarter assessment results into satisfactory and not satisfactory classes. A grid search is then implemented to determine the optimal values of LS features in the second quarter that maximize students' academic performance. Comparisons between the actual and optimal values of LS features in the second quarter are made based on defined thresholds and divided the learners into "Supported" and "Not Supported." Finally, the Mann-Whitney U test is used to compare the second quarter assessment grade between Supported and Not Supported groups. The findings depict that the statistical differences are significant, i.e., students in the Supported group achieved better grades in the second quarter assessment compared to the Not Supported group. Adopting an individual's learning style to the appropriate educational intervention can significantly impact student performance. The final contribution is to propose an approach to personalizing learning resources based on students' learning styles in an online learning environment. A student's learning style affects their learning attitude, satisfaction level, and academic performance. So, to maintain learners' interest and enhance their academic performance, it is crucial to consider their learning styles when developing e-learning systems. Students' interactions with the online learning environment are captured in a large amount of data during the learning process. The collected data can be used to analyze learners' behavior to determine their learning styles. In the proposed approach, the learning style and demographic features are then utilized for training machine learning models to predict students' academic performance in each quarter of different courses. The most accurate prediction model for each quarter is then used to find learning style features that maximize students' pass rates. We statistically prove that students whose actual learning style features were close enough to the ones calculated by the approach achieved better grades. To improve students' academic performance each quarter, we suggest two strategies based on the learning style features calculated by the process. [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.]