Discriminant Analysis in Bilinear Processes Using State Space Methodology

This paper is concerned with discriminating between two bilinear time series processes. The classical methods to discrimination makes stationary and linearity assumptions to process. However, in general, bilinear process are not stationary or linear. The method suggested in this paper is based on state space methodology and use Cumulative Sum Control Chart. Classification of series to H1 or H2 is is treated like a sequential likelihood ratio test where optimal decision to be made as soon as possible. The method has been carrid out to bilinear processes, but can be used to any adopted processes to state space models.