Time series discrimination using likelihood function of discrete wavelet coefficients

In this paper، the likelihood ratio of two normal density functions are approximated using discrete wavelet transform and a nonparametric criteria for discrimination of a stationary time series models in the wavelet domain are obtained. The performance of discriminant rule is shown in ARMA models using simulation techniques. In addition to do not requiring to parametric model، speed calculations for large time series and very low error rate are two characteristics of wavelet discriminate criteria. In this paper، the likelihood ratio of two normal density functions are approximated using discrete wavelet transform and a nonparametric criteria for discrimination of a stationary time series models in the wavelet domain are obtained. The performance of discriminant rule is shown in ARMA models using simulation techniques. In addition to do not requiring to parametric model، speed calculations for large time series and very low error rate are two characteristics of wavelet discriminate criteria.