Box-Jenkins models have some important limitations to the procedure : (a) They require a great deal of time, efforts and expertise for the model identification. (b) They require an extensive amount of past observations to identify an acceptable model. (c) The model selected is a constant model in time. Therefore, the Kalman Filter is recommended as a technique to overcome the three problems mentioned above. The research reported here uses the Kalman Filter algorithm to propose Kalman-AR(p) model. The data analysis shows that the Kalman-AR(p) model proposed can be used to resolve the problems of Box-Jenkins AR(p)model. It is seen that the Kalman Filter has great potentials for real-time industrial applications.