摘要
根据选择的嵌入维数,建立了基于支持向量回归的中国煤炭需求量预测模型.用1980—2002年的中国煤炭需求量构造了支持向量机的输入向量和输出向量;经过与线性核函数及Sig-moid核函数的对比,选用基于径向基函数(RBF)作为核函数,在分析预测误差和模型参数关系的基础上,选择了合适的参数;建立了多输入、单输出的支持向量机(SVM)预测模型.用检验样本与基于RBF神经网络模型的预测进行了比较,结果表明支持向量机模型在训练样本较少的情况下,仍有较高的预测精度和较强的泛化能力,证明了该模型对近期的预测是可靠的.最后用训练好的支持向量机模型很好地预测了2003-2006年我国的煤炭需求量.
A forecasting model for coal demand of China using a support vector regression was constructed. With the selected embedding dimension, the output vectors and input vectors were constructed based on the coal demand of China from 1980 to 2002. After compared with lineal kernel and Sigmoid kernel, a radial basis function(RBF) was adopted as the kernel function. By analyzing the relationship between the error margin of prediction and the model parameters, the proper parameters were chosen. The support vector machines (SVM) model with multi-input and single output Was proposed. Compared the predictor based on RBF neural networks with test datasets, the results show that the SVM predictor has higher precision and greater generalization ability. In the end, the coal demand from 2003 to 2006 is accurately forecasted.
出处
《中国矿业大学学报》
EI
CAS
CSCD
北大核心
2007年第1期107-110,共4页
Journal of China University of Mining & Technology
基金
国家自然科学基金项目(60575046)
关键词
支持向量机
煤炭需求量
预测
support vector machine
coal demand
prediction