摘要
针对选煤厂日用水量时间序列的预测问题,提出应用最小二乘支持向量机(LSSVM)这一新的机器学习方法来实现日用水量的短期预测.借鉴多层动态自适应优化算法的思想,提出最小二乘支持向量机参数优化的多层动态交叉验证法;用微熵率法求得选煤厂日用水量时间序列的最佳嵌入维数和最佳延迟参数,重构相空间,建立了基于最小二乘支持向量机的选煤厂日用水量时间序列等维信息一步预测模型.预测结果表明:基于LSSVM的预测模型的预测精度比BP神经网络预测模型的预测精度要高,能够满足选煤厂日用水量预测的需要.
Applied a novel machine learning algorithm-least squares support vector machines(LSSVM) into coal washery daily water consumption times series prediction.Firstly,referencing the principle of multi-layer adaptive best-fitting parameters search algorithm,a LSSVM's parameters optimization method which was called multi-layer dynamic cross-validation algorithm was proposed,and then the optimal embedding dimension and delay time were obtained by the differential entropy ratio method.In the reconstructed phase spa...
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2007年第10期1093-1097,共5页
Journal of China Coal Society
基金
江苏省自然科学基金资助项目(BK2003026)
关键词
最小二乘支持向量机
选煤厂日用水量
参数优化
BP神经网络
预测
least squares support vector machines
coal washery daily water consumption
parameters optimization
BP neural network
prediction
