摘要
针对风速随机性大、影响因素多、预测准确度不高的情况,基于支持向量机与信息几何的统计学关联性,从信息几何学角度分析核函数的几何结构,构造数据依赖核函数,并与支持向量机回归相结合,形成数据依赖核支持向量机回归(Data Dependent Kernel-SVR,DDK-SVR)方法.将该方法用于风速预测中,建立DDK-SVR风速预测模型,并将预测结果与传统支持向量机、神经网络方法进行对比.结果表明,DDK-SVR方法具有更高的预测精度.
Wind is random and has many factors. Besides,the prediction accuracy of wind is not high. Therefore,based on the statistics relationship between Support Vector Machine( SVM) and information geometry,the geometry of kernel function is analyzed. A data dependent kernel is constructed and combined with Support Vector Regression( SVR). Then,the support vector regression machine with data dependent kernel is proposed. We build a wind speed forecasting model and forecast the wind speed. Compared with SVM and neural networks,DDK-SVR method has higher prediction accuracy.
出处
《南京师大学报(自然科学版)》
CAS
CSCD
北大核心
2014年第3期15-20,共6页
Journal of Nanjing Normal University(Natural Science Edition)
基金
国家自然科学基金(61103141)
江苏省自然科学基金(BK2012858)
江苏省高校自然科学研究资助项目(13KJB520015)
关键词
风速预测
数据依赖核
支持向量机回归
wind speed forecasting
data dependent kernel
support vector regression machine