A Kalman filter based DSP method for prediction of seasonal financial time series with application to energy spot price prediction

In this work, energy spot price prediction is used to motivate a holistic signal processing approach to modeling and predicting nonstationary time series having a structure that is a mixture of quasi-periodic, cyclo-stationary, and locally regular stochastic components. The approach is iterative in the sense that the Kalman filter model used for estimation and prediction is repeatedly adjusted, based on exposure of hidden model structure identified using point spectrum and cyclo-stationary signal processing tools. It is shown that this holistic approach achieves reasonable 1-day and 7-day spot price prediction accuracy.