In this paper, we designed a 3

文中设计了一个3层径向基神经网络(RBFN)用于对企业的5项评价指标进行聚类分析，并与蚁群算法做了比较分析。RBFN由输入层 到隐含层采用传统的K一均值算法，隐含层到输出层通过“模2递减”学习速率的BP学习；蚁群算法根据信息素的分配能够自动调整收索 路径，从而达到数据自动聚类的目的。结果表明，与蚁群算法相比，改进RBFN具有快速收敛、自动识别奇异样本的优点，而蚁群算法 无须教师学习，并能够达到全局最优。-In this paper, we designed a 3-layer RBF neural network (RBFN) for the 5-to-business evaluation indicators cluster analysis and ant colony algorithm has done a comparative analysis. RBFN from input layer to hidden layer using the traditional K-means algorithm, hidden layer to output layer through the Mode 2 decreasing learning rate of BP learning ant colony algorithm based on pheromone can automatically adjust the allocation of land Faso path, thereby to achieve the purpose of automatic data clustering. The results showed that compared with the ant colony algorithm to improve the RBFN has a fast convergence, automatic identification of singular advantage of the sample, while the ant colony algorithm do not need teachers to learn and be able to reach the global optimum.

RBF.pdf