Adaptive cruise control (ACC) extends the conventional cruise control system by controlling the relative speed and distance to other vehicles. In order to design an ACC controller it is essential to have a model on drivers' behavior. Our approach to find the dynamical models of drivers' behavior is to use system identification. A basic data analysis was made by means of the system identification methodology, and several models of drivers' longitudinal behaviors are proposed, including both linear regression models and subspace based models. The detection of a driver changing his/her behavior in various situations to a deviant behavior is useful. For that purpose the GARCH model used to model the driver in arousal situations, where the driver changes his/her behavior, is proposed.