Evolutionary Wavelet Networks for Statistical Time Series Analysis

In this paper I describe an evolutionary wavelet network to optimize the filtering of a statisticaltime series into separate contributions. The wavelet base is regarded as a neural network where thenetwork nodes are discrete wavelet transforms, the wavelon, and the network structure and parameters areselected through evolutionary techniques. With this combined approach I can separate stochastic fromstructural components within an optimized framework and finally I can perform optimized predictiveanalysis on the time series components.