Information Technology Reference
In-Depth Information
Where, p is the input vector, tt is the target vector. SPREAD is the density of basis
functions, SPREAD is larger the function is smoother, where selecting SPREAD =
1.Then testing the neural network and verifying the prediction performance. MATLAB
codes are as follows:
y=sim(net,P_test)
Where, P_test is the network test samples. The results are as follows:
y= 0.3666 0.3887 0.4560 0.5024
After training and testing the network, the network output values obtained through
inverse transform are compared with the actual values to check whether it meet the
requirements of their error, as shown in table 2.
Table 2. RBF neural network analysis table accuracy
Year
Actual value(million)
Fittedvalue(million)
Absolute error
2005
1516.47
1516.40
0.07
2006
1605.02
1604.80
0.22
2007
1873.78
1874.00
-0.22
2008
2060.00
2059.6
0.4
We can see from Table 3, RBF neural network can reach 99.99% accuracy. it meets
the prediction requirements.This provides an accurate basis for predicting tourists
quantity in the future.
Enter the actual value from 2004 to 2008, we can obtain the normalized predicted
value in 2009.Similarly, after a multi-step iterative, we can get the normalized
predicted value from 2010 to 2018, as shown in table 3.
Table 3. The normalized predicted value of tourists quantity from 2009 to 2018
year
2009
2010
2011
2012
2013
Predictive value
0.58336
0.63658
0.67832
0.72896
0.76542
year
2014
2015
2016
2017
2018
Predictive value
0.81986
0.83942
0.84738
0.85106
0.86256
Then after inverse transform, the predicted value can be obtained from 2009 to 2018,
as shown in table 4.
Table 4. Predictive value of tourists quantity from 2009 to 2018
Year
2009
2010
2011
2012
2013
Predictivevalue(million)
2373.44
2586.32
2753.28
2955.84
3101.28
Year
2014
2015
2016
2017
2018
Predictive value(million)
3279.41
3397.68
3437.52
3442.24
3486.24
Search WWH ::




Custom Search