A new approach to modeling of stochastic processes based on principal components of Hankel matrix

This paper describes a new method of modeling of stochastic processes by simultaneously determining the order and parameters of model. Process information is contained in observed time series-the system output or covariation of the output. The new method is based on singular value decomposition of Hankel matrix of covariation of process output. It differs from a previous similar approach in a more adequate definition of optimization problem. Noniterative algorithm of time series modeling is realized. The capabilities of the algorithm are illustrated through an example and are compared to the known approach.