16 April 2014 Initial parameters problem of WNN based on particle swarm optimization
Author Affiliations +
Proceedings Volume 9159, Sixth International Conference on Digital Image Processing (ICDIP 2014); 91590V (2014) https://doi.org/10.1117/12.2064241
Event: Sixth International Conference on Digital Image Processing, 2014, Athens, Greece
Abstract
The stock price prediction by the wavelet neural network is about minimizing RMSE by adjusting the parameters of initial values of network, training data percentage, and the threshold value in order to predict the fluctuation of stock price in two weeks. The objective of this dissertation is to reduce the number of parameters to be adjusted for achieving the minimization of RMSE. There are three kinds of parameters of initial value of network: w , t , and d . The optimization of these three parameters will be conducted by the Particle Swarm Optimization method, and comparison will be made with the performance of original program, proving that RMSE can be even less than the one before the optimization. It has also been shown in this dissertation that there is no need for adjusting training data percentage and threshold value for 68% of the stocks when the training data percentage is set at 10% and the threshold value is set at 0.01.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Chi-I Yang, Kaicheng Wang, Kueifang Chang, "Initial parameters problem of WNN based on particle swarm optimization", Proc. SPIE 9159, Sixth International Conference on Digital Image Processing (ICDIP 2014), 91590V (16 April 2014); doi: 10.1117/12.2064241; https://doi.org/10.1117/12.2064241
PROCEEDINGS
4 PAGES


SHARE
Back to Top