Please use this identifier to cite or link to this item: http://hdl.handle.net/1959.14/148096
64 Visitors72 Hits0 Downloads
A Stratified model for short-term prediction of time series
Pacific Rim International Conference on Artificial Intelligence (11th : 2010) (30 August - 02 September 2010 : Daegu, Korea)
Zhang, Byoung-Tak and Orgun, Mehmet A.. PRICAI 2010 : trends in artificial intelligence : 11th Pacific Rim International Conference on Artificial Intelligence, Daegu, Korea, August 30-September 2, 2010 : proceedings, p.372-383
This paper develops a model for short-term prediction of time series based on Element Oriented Analysis (EOA). The EOA model represents nonlinear changes in a time series as strata and uses these in developing a predictive model. The strata features used by the EOA model have the potential to improve its forecasting performance on non-linear data relative to the performance of existing methods. We demonstrate the characteristics of the EOA model using an empirical study of stock indices from eight major stock markets. The study provides comparisons of the accuracy and time efficiency between ARIMA, Neural Networks and the EOA model. Our findings indicate that the EOA model is a promising approach for short-term time series prediction.