Understanding a building's ambiance and user's preferences and then providing corresponding comfort is substantial in a smart home environment. In this work, we aim to design a model for an intelligent system (building) controller whose intelligence is adaptation, in changing situations, according to the preferences of occupants without their intervention. Adaptation, according to Humphreys is: "If a change occurs such as to produce discomfort, people react in ways which tend to restore their comfort". Therefore, our adaptive system restores comfort to the occupants based on their preferences, it has the ability to self-regulate and adapt to the climate conditions in buildings. In order to use adaptive control a model of the building is necessary and predictive control is very important because it includes a model for future disturbances. The starting point of this work was the modeling of multisensory comfort, and, the dynamic and adaptive behavior of an occupant with his environment. A key element was to find a way to model these adaptive actions. To achieve our goal, we used Bayesian networks that are powerful tools for decision and reasoning under uncertainty.