Abstract
Wavelet transform has been applied to many areas. In this paper, the application of wavelet transform denoising to financial time series is considered. Daubechies wavelet was used to denoise the NASDAQ-100 index with different noise models. The goodness of these denoising techniques was evaluated using the variance of prediction error. In-sample testing and out-of-sample testing were studied. It was shown that the prediction error of the NASDAQ-100 index is substantially reduced when it is denoised. The reduction is maximum, when colored noise assumption and universal thresholding are used.