Abstract: BP (back-propagation, BP) algorithm is a way widely used in artificial neural networks. However, this algorithm exposes its defects in tardiness of convergence rate and easily falling into partial minimum. Thus,aiming at those defects,measures of improved BP algorithm are adopted by additional momentum and variable learning rate, and the simulation of MATLAB to Guangxis GDP from 2001 to 2010 shows that the multiple BP algorithm is superior to the standard BP algorithm in both convergence rate and accuracy.

Key words: Back-Propagation algorithm, Artificial neural networks, Improved Back-Propagation algorithm, MATLAB