Statistical and Soft Computing Methods Applied to High Frequency Data
Dusan Marcek and Alexandra Kotillova
We evaluate statistical and machine learning methods for predicting different high frequency data sets. Firstly, in this paper we develop forecasting models based on the statistical (stochastic) methods, and on the soft methods using neural networks for the time series of daily exchange rates AUD currency against US dollar. Secondly, we evaluate statistical and machine learning methods for half-hourly 1-step-ahead electricity demand prediction using Australian electricity data. To illustrate the forecasting performance of these approaches the learning aspects of RBF networks are presented. We also show that an RBF neural network trained by genetic algorithm can achieve better prediction result than classic one. It is also found that the risk estimation process based on soft methods is simplified and less critical to the question whether the data is true crisp or white noise.
Keywords: ARIMA and ARCH/GARCH models, information granules, neural networks, support vector regression, genetic algorithms, forecast accuracy, half-hourly electricity demand prediction.