References in periodicals archive ?
Then, a combination of wavelet transform and Artificial Neural Network (ANN), that is, Wavelet Neural Network (WNN), is used to predict the network from the quantitative aspect.
In the quantitative prediction aspect, we use WNN for time-series prediction.
g(x) is our prediction function and is represented by WNN with its hidden layer of the Morlet wavelet.
WNN is then used to predict the future values of the four measures from the quantitative perspective.
After running 100 times with WNN and baseline methods, the average values of the prediction results are obtained.
As the WNN has the best prediction performance, we then apply the WNN to quantitatively predict the measures in the subsequent four timestamps.
From Figure 9, we observe that the prediction values of the four measures by WNN are consistent with the trend we inferred qualitatively in Table 3.
On the other hand, WNN can quantitatively predict the future trends based on the wavelet transform.
Caption: Figure 9: Prediction results of four measures by WNN for the subsequent four timestamps.
For more information on the event, visit Giuffre Volvo's Facebook at www.facebook.com/GiuffreVolvo or the 98.7 WNNS site at www.wnns.com/page/two-week-test-drive.
Following graphs show the error versus number of iterations and graph of test signal when applied to WNN for different wavelets and fixed number of iterations=1600
Acronyms browser ?
Full browser ?