Application of RBF Neural Network Model on Forecasting Tourists Quantity in Hainan Province - Advanced Research on Computer Education, Simulation and Modeling - page 308

output changes is directly related to the performance of neural network prediction. The
number of tourist is restricted by many factors, for example, geography, environment,
culture, government policy, etc. If all these factors are considered, it will bring a lot of
inconvenience to predict. Tourists quantity every five years as the neural network input
variables is the innovation of the article, so Input samples can be determined by the
input variable. We can select the sixth year number of tourists after every five years as
the neural network output variable.

2.2 Input Samples Pretreatment

Since the implicit function of RBF neural network is Gaussian function, which general
requires for input value between 0 and 1, do normalize on the number of Hainan
province tourists from 1988 to 2008. Normalization is basically the same way to
statistical data normalization, generally using the following form:

X

X −

__
X =

min

(5)

X X

−

max min

Where X is the actual value of sample;

X takes a large value, ensuring forecast
year is less than the value; mi X takes a sample of data is less than the minimum value
to ensure normalized value is not close to 0. After the pretreatment of data completes
the training, do process data (inverse transform) to get the actual value.

max

Table 1. The actual number of visitors of Hainan Province in 1988 to 2008

Years 1988 1989 1990 1991 1992 1993 1994

Tourists quantity(million) 118.54 88.05 113.46 140.61 247.37 274.41 289.60

Year 1995 1996 1997 1998 1999 2000 2001

Tourists quantity(million) 361.01 485.82 791.00 855.97 929.07 1000.76 1124.76

Year 2002 2003 2004 2005 2006 2007 2008

Tourists quantity(million) 1254.54 1234.11 1402.88 1516.47 1605.02 1873.78 2060.00

Note: Table 1 Data from the Hainan Provincial Bureau of Statistics.

2.3 Determining Training Samples and Test Samples

From the above we can determine the number of input neuron of RBF neural network is
5, and the number of output neurons is 1. Treating the samples as follows [6]: Input
neuron P=[p(t-5),p(t-4),p(t-3),p(t-2),p(t-1)]; Output neurons T=[p'(t)]. Where, t =
1993, 1994 ... ... 2008, P (t) denote the normalized number of tourism at t year. In this
method, we can obtain the training samples and test samples.