Research of BP Algorithm Based on Fusion Technique

The BP neural network algorithm data can be parallel processing, information processing capability is strong, itself is learning, association and memory capacity, avoids the limitations of traditional methods and the subjective and arbitrary of expert evaluation, the source of a single cause data with evaluation objects evaluation model is not between objective simplified, and a single source of data led to the not objective simplification between evaluation model and evaluation. But it also has the network training time is too long, easily falling into local minima, cann´t training and other shortcomings. In this paper, we design a algorithm with principal component analysis, particle swarm optimization algorithm and BP neural network. The new algorithm has well application ability, and compared with the BP algorithm, it has small errors and short training time.