Title :
Discussion on the Relation Between SVM Training Sample Size and Correct Forecast Ratio for Simulation Experiment Results
Author :
Zhu, Shuguang ; Zhu, Fangzhou ; Fan, Weibing ; Mo, Qian ; Li, Zhiqiang ; Hu, Xiaofeng ; Si, Guangya
Author_Institution :
Center for Eng. Design & Res., Headquarters of Gen. Equip., Beijing, China
Abstract :
A series of support vector machine (SVM) forecast experiments are carried out to reveal the relation between the SVM training sample size and SVM correct forecast ratio for simulation experiment results. Experiment results show that the SVM correct forecast ratio increases to some extent with the number of training samples becoming more and then keeps unchanged even if the SVM training sample number increases further. And SVM has also been proved to be able to overcome the over-fitting issue always afflicting Back-Propagation Neural Networks (BPNN).
Keywords :
backpropagation; neural nets; support vector machines; SVM training sample size; backpropagation neural networks; forecast ratio; simulation experiment; Computer networks; Computer simulation; Design engineering; Intelligent networks; Predictive models; Sampling methods; Support vector machine classification; Support vector machines; Technology forecasting; Testing; correct forecast ratio; forecast experiment; support vector machine; training sample size;
Conference_Titel :
Intelligent Computation Technology and Automation (ICICTA), 2010 International Conference on
Conference_Location :
Changsha
Print_ISBN :
978-1-4244-7279-6
Electronic_ISBN :
978-1-4244-7280-2
DOI :
10.1109/ICICTA.2010.301