Generalized Regression Neural Network for Forecasting Time Series with Multiple Seasonal Cycles
This paper presents a method of forecasting time series with multiple seasonal cycles based on Generalized Regression Neural Network. This is a memory-based, fast learned and easy tuned type of neural network. The time series is preprocessed to define input and output patterns of seasonal cycles, which simplifies the forecasting problem. The method is useful for forecasting nonstationary time series with multiple seasonal cycles and trend. The model learns with the help of differential evolution or simple enumerative method. The performance of the proposed method is compared with that of other forecasting methods based on Nadaraya-Watson estimator, neural networks, ARIMA and exponential smoothing. Application examples confirm valuable properties of the proposed method and its highest accuracy among the methods considered.
KeywordsSeasonal time series forecasting generalized regression neural network differential evolution pattern similarity based forecasting short-term load forecasting
Unable to display preview. Download preview PDF.
- 7.Dudek, G.: Forecasting Time Series with Multiple Seasonal Cycles using Neural Networks with Local Learning. In: Rutkowski, L., Korytkowski, M., Scherer, R., Tadeusiewicz, R., Zadeh, L.A., Zurada, J.M. (eds.) ICAISC 2013, Part I. LNCS, vol. 7894, pp. 52–63. Springer, Heidelberg (2013)CrossRefGoogle Scholar
- 8.Hyndman, R.J., Khandakar, Y.: Automatic Time Series Forecasting: The Forecast Package for R. Journal of Statistical Software 27(3), 1–22 (2008)Google Scholar