This article presents the results of an experiment in personalizing course content and learning activity model tailored for online courses based on students' learning styles. The main research objectives are to design and pilot a model to determine students' learning styles to create personalized online courses. The study also addressed an effective method to identify learning styles and evaluate how student's learning styles impact students' learning outcomes. With an aim to personalize suitable content and learning process for each student, machine learning techniques have been used to detect students' learning styles and classified them into learning styles based on the VARK model by analyzing learning activity data. Based on students' learning styles, rules were proposed to select appropriate content and learning processes. The research results show that the SVM method performs the best among classification methods used to determine students' learning styles. In addition, a plugin was developed on the Moodle system to support the automatic identification of students' learning styles, based on which a learning process and appropriate content were generated to test the model's results. The experiment results also indicate that students with a visual learning style had better learning outcomes in theory-oriented courses. In contrast, students with a kinesthetic learning style had better learning outcomes in practice-oriented courses. Although the experiment was only conducted on a small scale, the positive results show that the model can fully meet the needs of large-scale LMS systems.