THE SINGULAR-VALUE DECOMPOSITION - APPLIED IN THE MODELING AND PREDICTION OF QUASI-PERIODIC PROCESSES

Singular value decomposition is used as a tool in the analysis of quasiperiodic processes. The prime orthogonal components are used to separate the regular and the irregular parts of the process. Linear prediction of the orthogonal components of the regular part leads to periodic prediction. For quasiperiodic processes, the same on the non-linearly transformed irregular part is used as additional component to produce periodic prediction. Prediction using singular value decomposition is compared numerically with that obtained using some established methods and the superiority of the former is demonstrated.