SPECTRAL REGULARIZATION, DATA COMPLEXITY AND AGENT BEHAVIOR

The simple and efficient procedure of trend identification in economic time series using the FFT method has been proposed. The effect of trend identification on the behavior of agents on stock markets has been considered. The spectral decomposition of economic data as ill-posed problems has been studied. Connection of trend identification with data smoothing and regularization for numerical differentiation of empirical data has been discussed. Relations among data subjective complexity, spectral regularization parameter and temporal preferences of agent have been shown. The selection of the cut-off frequency for data smoothing should correspond to the investment horizon of the economic agent.