Research interest in the field of adaptive systems is steadily increasing over the past 30 years, which is also true for the field of applications that are actively configured to support users. The field of e-Learning is such an area and it has rapidly evolved into a viable alternative to the traditional learning environment, so it has become the focus of research on adaptive systems. Despite attention that adaptive learning has recently received focus as a social process, learning is being treated as a process consisting of interconnected activities but not consumption (passive or active) learning content, to an individual or group level. The premise of the line of work reported in this paper is that the monitoring and interpretation of online learning activities can eventually lead to a rich model for learner users, which in turn enables new forms And expanded the adaptive response in the context of e – learning. In this article, we propose a new approach to deal with extraction, analysis and interpretation of information from sequential user activities of this information as the ultimate goal to derive intention knowledge adaptation of natural learning behavior. As a proposed approach, we use models obtained in combination with other types of monitoring data, especially for modeling and discovering user activity sequences (like Markov model discrete) Integrate into adaptation cycles (Including discovery of new semantics of learner ‘s behavior) semantically meaningful information of the learner. Discovery is guided in both cases by grouping learning activity sequence patterns. Grouping can be applied at three levels according to the purpose of discovery. To demonstrate the feasibility of the approach, we applied real-world data to the domain of problem solving. Problem solving is an important part of the learning process in traditional learning environments and e-learning. Learners’ problem solving styles are on different levels but they are explained in the same way as the most well-known learning style, and in most cases. To solve problems, atoms are applied to problems of different levels of complexity of complexity. While analyzing Behavior for problem solving is done using mostly statistical knowledge to study styles based on user activity data, which is almost exactly investigated and cannot be modeled without more detailed sequence information when appropriate for the proposed approach. For the purposes of the work described in this document, we have developed a custom template to capture the sequence of student activities involved in problem solving in a particular intellectual education system (ITS). We set up a resolution sequence student problem that uses these models, depending on the above three levels, then for clustering; (a) Detects the resolution of the style of predefined problem (s). (B) learn a new style of problem solving according to predefined learning dimensions. And (c) finds a potentially interesting learning dimension and associated problem solving style.