Forecasting share price using wavelet transform and LS-SVM based on chaos theory

• Published in: 2008 IEEE Conference on Cybernetics and Intelligent Systems pp. 300 - 304
• Main Authors: Jianguo Zhou, Tao Bai, Aiguang Zhang, Jiming Tian
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2008
• Subjects: Accuracy; Ambient intelligence; Chaos; Chaotic Time Series; Least squares methods; LS-SVM; Mutual information; Predictive models; Share Price Forecast; Share prices; Support vector machines; Time series analysis; Wavelet Transform; Wavelet transforms
• ISBN: 1424416736; 9781424416738
• ISSN: 2326-8123
• DOI: 10.1109/ICCIS.2008.4670953

In the analysis of predicting share price based on least squares support vector machine (LS-SVM), the instability of the time series could lead to decrease of prediction accuracy. On the other hand, three SVM parameters, c, epsiv and sigma, must be carefully predetermined in establishing an efficient LS-SVM model. In order to solve the problems mentioned above, in this paper, the hybrid of wavelet transform (WT) with LS-SVM model was established. First the chaotic feature of share price is verified with chaos theory. It can be seen that share price possessed chaotic features, providing a basis for performing short-term forecast of share price with the help of chaos theory. Average mutual information (AMI) method is used to find the optimal time lag. Then the time series is decomposed by wavelet transform to eliminate the instability. Caopsilas method is adopted to determine free parameters of support vector machines. Additionally, the proposed model was tested on the prediction of share price of one listed company in China. Especially, In order to validate the rationality of chosen dimension, the other three random dimensions were selected to compare with the calculated dimension. And to prove the effectiveness of the model, PSVM algorithm was used to compare with the result of WT-SVM. Experimental results showed that the proposed model performed the best predictive accuracy and generalization, implying that integrating the wavelet transform with LS-SVM model can serve as a promising alternative for share price prediction.