Pervasive computing (also referred to as ubiquitous computing or ambient intelligence) aims to create environments where computers are invisibly and seamlessly integrated and connected into our everyday environment. Pervasive computing applications are often interaction transparent, context aware, and experience capture and reuse capable. Interaction transparency means that the human user is not aware that there is a computer embedded in the tool or device that he or she is using. Context awareness means that applications and services should be aware of their contexts and automatically adapt to their changing contexts-known as context-awareness. An experience capture and reuse capable application can remember when, where, and why something was done and can use that information as input to solve new tasks. Context, as defined by Anind K. Dey in 2001, is any information that can be used to characterize the situation of an entity. An entity is a person, place, or object that is considered relevant to the interaction between a user and an application, including location, time, activities, and the preferences of each entity. A system is context-aware if it can extract, interpret and use context information and adapt its functionalities to the current context of use. In a pervasive environment, several middleware have been proposed to manage context information such as the Context Toolkit, Aura, Amigo, etc.... In these environments, Context-aware applications adapt their behaviour using Event-Condition-Action (ECA) rules, also referred to as adaptation rules. An ECA rule defines an action to be performed as a reaction to some event under a certain condition. The event part refers to context changes, the condition part to the current context and the action part to an adaptive behaviour. After studying several adaptation works in a pervasive environment we can conclude that the adaptation conducted until now is only supported by static rules in a way that they were specifi- d by developers during the implementation phase and the applications behavior was restricted to the actions specified in that rules. However, in such an environment, some of the adaptation rules change naturally over time, particularly those related to the preferences of a human user. Indeed, the later can decide to perform a particular action in a specific context at a given time and act differently against the same context at a later time. Consequently, the predefined static adaptation rules can't support this frequent change of users' preferences and thus, it is difficult to provide the users with the automatic personalized services. Hence, our work aims to propose an approach for adaptation to the dynamic changing context in pervasive environments. The principle of our approach is to distinguish static and dynamic adaptation rules based on the user preferences. We use the history of users interactions to predict their future behavior and thus to define new adaptation rules or update the existing ones. Our solution combines machine learning techniques, specifically the Case Based Reasoning method (CBR) and data mining techniques to extract and update rules. We are actually at the stage of implementing our approach which we consider to be giving a new impetus to research in the field of adaptation and personalization in pervasive environments.