AI (artificial intelligence) techniques, especially neural network, have been used widely for business forecasting. There are so many factors affecting the business forecasting that the input nodes number is large. Conventional neural network methods suffer from limitations, which make them less than adequate for decision making in dynamic business environment. In order to reduce the input nodes, the factors that affect the business forecasting are standardized firstly. Then they are reduced using the principle components analysis method. For hidden nodes, their number is firstly limited to less than the square root of product of input nodes number and output nodes number. Then the correlation coefficients between different hidden nodes in same layer are calculated. Lastly the hidden nodes are merged or deleted according to correlation coefficients. The structure of improved BP neural network (IBNN) is optimized by above method. The result of the business forecasting using the IBNN is shown to be satisfying
[1]
Rong Li.
A Method to Determine the Structure and Parameters of BP Neural Network from Knowledge
,
2003
.
[2]
Etienne Barnard,et al.
Optimization for training neural nets
,
1992,
IEEE Trans. Neural Networks.
[3]
Bimal K. Bose.
Neural Network and Applications
,
2006
.
[4]
Bo Yang,et al.
BP neural network optimization based on an improved genetic algorithm
,
2002,
Proceedings. International Conference on Machine Learning and Cybernetics.
[5]
Geoffrey E. Hinton,et al.
Learning internal representations by error propagation
,
1986
.
[6]
Jude W. Shavlik,et al.
Extracting Refined Rules from Knowledge-Based Neural Networks
,
1993,
Machine Learning.