摘要
针对传统的GPS高程拟合方法要求有足够多样本数据的缺陷,本文采用粒子群(PSO)算法优化最小二乘支持向量机(LSSVM)参数的方法进行GPS高程拟合。实验表明,在有限样本的情况下,PSO-LSS-VM模型不仅发挥了LSSVM处理小样本数据的能力,而且通过PSO优化后的LSSVM能够选择出合适的参数;与LM-BP神经网络、标准最小二乘支持向量机等方法比较,PSO-LSSVM模型拟合精度较高。
Viewing from the defects that the traditional GPS elevation fitting methods require enough data,the paper put forward that using particle swarm optimization(PSO) algorithm to optimize least squares support vector machines(LSSVM) parameter for GPS elevation fitting.Experiments showed that,in the cases of limited data,not only PSO-LSSVM model could bring into play LSSVM ability to process the small sampled data,but also LSSVM optimized by PSO could choose the appropriate parameters.Comparing with LM-BP neural network and standard least squares support vector machine,PSO-LSSVM model has better fitting accuracy.
出处
《测绘科学》
CSCD
北大核心
2010年第5期190-192,共3页
Science of Surveying and Mapping