改进海鸥算法优化BP神经网络的GPS高程拟合方法  

GPS Elevation Fitting Method Using Improved the Seagull Algorithm to Optimize BP Neural Network

作　　者：伍勇 WU Yong(Hubei Province Jingzhou Hydrology andWater Resources Survey Bureau,Jingzhou 434000,China)

机构地区：[1]湖北省荆州市水文水资源勘测局,湖北荆州434000

出　　处：《地理空间信息》2025年第2期108-111,共4页Geospatial Information

摘　　要：针对BP神经网络受参数初始权值和阈值影响高程拟合精度问题,改进海鸥算法弥补BP神经网络缺陷,使拟合模型得到更优参数。将海鸥算法加入混沌映射初始种群和改变惯性权重运动方式,提高算法寻优精度、收敛速度且具有鲁棒性,利用改进海鸥算法的最优适应度个体神经元连接权值和阈值,以此建立改进海鸥算法优化BP神经网络的GPS高程异常值拟合预测模型,通过实际工程GPS高程数据对模型进行检核。结果表明提出改进优化模型比传统BP神经网络模型拟合精准性、稳定性高,对数据适应度更好,是一种可靠拟合模型。The initial weight and threshold of parameters will affect the elevation fitting of BP neural network,and the improved seagull algo-rithm can make up the defects of BP neural network,so that the fitting model can get better parameters.In this paper,we added seagull algorithm into the initial population of chaotic mapping and changed the inertia weights movement to improve the algorithm’s accuracy,convergence speed and robustness.We used the optimal adaptive individual neuron connection weights and thresholds of improved seagull algorithm to establish the prediction model of GPS elevation anomalies by using improved seagull algorithm to optimize BP neural network,and checked the model by the actual engineering GPS elevation data.The model is checked by actual engineering GPS elevation data.The results show that the improved mod-el proposed in this paper has higher fitting accuracy and stability than the traditional BP neural network model,and has better adaptability to the data,which is a reliable fitting model.

关 键 词：海鸥算法 混沌映射 BP神经网络 高程拟合 

分 类 号：P228[天文地球—大地测量学与测量工程]