BP neural network evaluation method can overcome the shortcomings of traditional expert dependence, but the efficiency of evaluation will decline with the expansion of index system and fall into the local maximum value. Attribute reduction based on neighborhood rough set (RS) can optimize the structure and weight of BP neural network, but only considering the influence of a single attribute in importance calculation leads to index weight zeroing and attribute error deletion. For this reason, a grey correlation method was used to improve attribute reduction, and momentum factor was used to modify BP weight updating strategy, and a grey correlation RS-BP performance evaluation model was proposed. The simulation results show that the reduction speed and evaluation accuracy of the grey relational RS-BP neural network model are improved.
Discussion(0)
No comments yet. Be the first to comment.