论文部分内容阅读
利用遗传-蚁群混合算法(GAAA),对RBF神经网络的主要结构参数中心矢量、基宽向量和网络权重进行组合优化,建立了GAAA-RBF神经网络组合算法的工程估价模型.将55个工程造价案例,随机抽取10个作为预测样本,剩下的45个作为训练样本.通过与相同结构的RBF神经网络相比较,结果表明算法克服了RBF神经网络易陷于局部极值、搜索质量差和精度不高的缺点,改善了RBF神经网络的泛化能力,收敛速度快,输出稳定性好,提高了工程造价的预测精度.
RBF neural network’s main structural parameter center vector, base width vector and network weight are optimized and combined by genetic-ant colony hybrid algorithm (GAAA), and a project evaluation model of GAAA-RBF neural network combination algorithm is established.With 55 projects Cost case, 10 samples were randomly selected and the remaining 45 were used as training samples.Compared with the RBF neural network with the same structure, the results show that the algorithm overcomes the RBF neural network easily trapped in the local extreme, poor search quality and accuracy Not high shortcomings, improve the RBF neural network’s generalization ability, fast convergence, output stability is good, and improve the prediction accuracy of project cost.