Application of RBF Neural Network Model on Forecasting Tourists Quantity in Hainan Province - Advanced Research on Computer Education, Simulation and Modeling - page 307

The distance between the weight vector 1 W , connected to the inputting layer and
in the every neuron in the hidden layer, and the input vector is multiplied by the

X

q

threshold ,which is considered as its own input. As figure2 shows:

b 1 i

Fig. 2. The input and output of RBF neural network hidden layer neurons

Thus we get the ith input of hidden layer neuron which can be expressed as

k

q

:

i

( )

2

q

∑

q

k w 1 x b 1

= − ×

(2)

i j i j i

j

q

The ith output of hidden layer neuron can be expressed as

r

:

⎛ ⎞

( )

(3)

2

q

∑ wx 1

q

b 1

r

= exp − ×

⎜ ⎟

⎜ ⎟

i ji j i

⎝ ⎠

j

The output of RBF neural network is the weighted summation of each hidden layer
neurons's output and the excitation function is using pure linear function, so the qth
output layer neurons's output which is corresponding to the qth input can be ex-
pressed as

q

y

:

n

(4)

∑

q q

y = r × w 2 h

i i

i = 1

1.3 The Learning Algorithm of RBF Neural Network

RBF neural network learning process can be divided into two stages [5]: first stage,
self-organizing learning phase, this phase is the unsupervised learning process, solving
the center and variance of the hidden layer base functions; second stage, tutor learning
phase, this phase is solving weights which is between the hidden layer and output layer.

2 The Construction and Forecast of RBF Model on Forecasting
Tourists Quantity

2.1 RBF Neural Network Input Variables and Output Variables

Input variable selection is an important task before the RBF neural network modeling,
whether to choose a set of input variables which can best reflect the reason for desired