DISCRETE REPRESENTATION OF NONLINEAR PROCESSES USING TIME SERIES DATA.

This paper presents a modeling technique for the discrete representation of nonlinear processes. The modeling technique is referred to as the Nonlinear Data Dependent System (NLDDS) which contains a nonlinear pattern classification procedure and a model search algorithm. The method uses some statistical and graphic techniques to generate various linear and nonlinear patterns for data classification. System dynamics are characterized in terms of nonlinear autoregressive moving average (NLARMA) models.