摘要
Abstract
In response to the limitations of the back propagation neural network(BP)in elevation fitting,this paper proposed an improved artificial bee colony(ABC)algorithm for optimizing BP(MABC-BP)to enhance its convergence speed and fit-ting accuracy.The traditional ABC-BP method faces issues such as insufficient diversity in the initial population,premature convergence,and low search efficiency.In this study,Chebyshev chaotic mapping,global optimal solution introduction,and Levy flight methods were utilized to improve the algorithm.The fitting performance of BP,ABC-BP,and MABC-BP models was compared based on two different terrain characteristics.Experimental results show that for the Mine Area 1 data-set,the internal and external fitting accuracy of MABC-BP has increased by 57.58%and 53.99%,respectively,compared to BP,and by 46.38%and 40.48%,respectively,compared to ABC-BP.For the Mine Area 2 dataset,MABC-BP improves internal and external fitting accuracy by 65.74%and 64.31%,respectively,compared to BP,and by 53.08%and 50.47%,respectively,compared to ABC-BP.关键词
反向传播神经网络/高程拟合/人工蜂群算法/切比雪夫混沌映射/莱维飞行Key words
back propagation neural network/elevation fitting/artificial bee colony algorithm/Chebyshev chaotic mapping/Levy flight分类
测绘与仪器
