论文部分内容阅读
提出了一种在山区能够准确、稳定地预测未采样点土壤重金属浓度的集成径向基函数神经网络空间插值方法(IRBFANNs).该方法集成径向基函数神经网络和神经网络集成技术的优点.为了研究所提IRBFANNs方法的性能,进行了3组不同采样密度条件下的实验.通过M n元素插值的均方根误差和分布估计图对IRBFANNs和其他6个插值方法进行了比较.实验结果表明:IRBFANNs方法在精确性和稳定性方面优于其他参评方法,且在采样密度稀疏条件下该方法能够提供细节较丰富的分布估计图.
An integrated radial basis function neural network spatial interpolation method (IRBFANNs) is proposed to accurately and stably predict the concentration of heavy metals in unsampled soils in mountainous areas. This method integrates the advantages of radial basis function neural networks and neural network ensemble techniques. In order to study the performance of the proposed IRBFANNs method, three groups of experiments at different sampling densities were carried out.The IRBFANNs and other six interpolation methods were compared through the root mean square error and distribution estimation of M n element interpolation.The experimental results show that : The IRBFANNs method outperforms other methods in terms of accuracy and stability, and provides a more detailed distribution estimation map with less sample density.