论文信息 - A neural network based time series forecasting system

A neural network based time series forecasting system

Neural network based system is proposed as a tool supporting time series forecasting. Wavelets analysis is suggested for raw data preprocessing. The preprocessed data are fed into SOM neural networks for clustering. The system has been tested as a tool supporting stock market investment decisions and shows good prediction accuracy of the method. However effectiveness, measured by income from shares trading according to the system, is not satisfactory to the same extend.

Bohdan Kozarzewski

[1] Emanuele Canegrati. A Non-Random Walk down Canary Wharf , 2008 .

[2] O. Rioul,et al. Wavelets and signal processing , 1991, IEEE Signal Processing Magazine.

[3] Teuvo Kohonen,et al. Self-Organizing Maps , 2010 .

[4] Charles L. Wilson,et al. Self-organizing neural network system for trading common stocks , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[5] Janusz Morajda,et al. Application of Cluster Analysis Performed by SOM Neural Network to the Creation of Financial Transaction Strategies , 2005 .