Classification algorithm based on neural network and rough set
一种基于粗糙集神经网络的分类算法*

When neural network has high dimensional inputs, it may have a complex structure and too large systems, and also it may lead to slow convergence. In order to overcame this shortcoming it proposed a neural network based on decision rules (RDRN). It used rough set theory to get the most simple decision rules from the data samples, and then constructed a not fully connected neural network by the semantics of the decision rules. According to the semantics of decision-making rules, it calculated and initialized the network parameters to reduce the number of iterations of network training and to improve the convergence speed. At the same time, the ant colony optimization algorithm was used to find the optimal discrete value of continuous attributes of the net inputs in order to obtain an optimal network structure. Finally, experimental results of the proposed method, the traditional neural network methods and support vector classification methods were compared. Comparison showed that the neural network converges faster and had advantages of more efficient classification.