摘要
针对BP神经网络预测极易陷入局部最优解,利用思维进化算法优化BP神经网络的初始权值和阈值,提出基于思维进化法优化BP神经网络(MEA-BP)大坝变形预测模型。通过算例验证,并与BP神经网络、GA-BP神经网络对比分析表明,该模型能够克服多数进化算法问题及缺陷,同时避免遗传算法中交叉和变异算子双重性,提高算法的整体搜索效率,在一定程度上保证较优的局部预测值和较好的全局预测精度,具备快速收敛能力,验证了提出的MEA-BP神经网络预测模型在大坝变形预测中的可行性和实用性。
Aiming at the BP neural network prediction method to easily fall into the local optimal solution,a BP neural network based on the mind evolutionary algorithm(MEA-BP)dam deformation prediction model is proposed,which can optimize the BP neural network’s initial weight and threshold value.It is shown that the model can overcome the problems and shortcomings of most evolutionary algorithms,and avoid the duality of crossover and mutation operator in genetic algorithm,and improve the overall search efficiency of the algorithm by comparing the examples with the BP neural network and GA-BP neural network,which can guarantee the better local forecasting value and better global forecast precision,and has the ability of fast convergence.It verifies the feasibility and practicability of the proposed MEA-BP neural network prediction model in dam deformation prediction.
出处
《北京测绘》
2017年第3期75-78,91,共5页
Beijing Surveying and Mapping
关键词
思维进化法
BP神经网络
大坝变形预测
精度评定
mind evolutionary algorithm
BP neural network
dam deformation prediction
precision evaluation