This paper investigates the development of learner personas to enhance personalized learning in e-learning environments, focusing on student activity data from the educational recommender system ELARS. The presented study analyses student activity patterns from across three university STEM courses, where tasks were assigned for each course topic and made available for a 15-day period, during which students could complete them at their own pace. The data were analyzed using the K-Means clustering algorithm, and three key variables that reflect student characteristics were identified: (1) the number of completed tasks, which indicates differences in student engagement and persistence; (2) the timing of task completion within the 15-day period, highlighting various study habits, such as early completion, evenly distributed efforts, or procrastination; and (3) the preferred time of day for task engagement, distinguishing between daytime and night-time study routines. These insights provide a basis for creating learner personas, which, when combined with student model, can enable providing tailored feedback, resource recommendations, adaptive learning paths, and similar personalized learning interventions. Future work will involve validating these findings with a larger sample and developing recommendation algorithms to personalize learning experiences, particularly for STEM subjects. The research described in this paper contributes to the field of personalized education by identifying student habits, routines, and preferences that can be leveraged to meet diverse learning needs in e learning systems.