摘要
针对产品销售时序具有多维、小样本、非线性、随机性等特征,已有的支持向量核不可能精确逼近任意的销售时序曲线.将小波理论应用于支持向量核函数,并对标准支持向量机进行修正,形成一种新的小波支持向量机(WN-ν-SVM).设计了自适应正态变异粒子群算法(ANPSO)对小波支持向量机模型参数进行辩识,并进行了汽车销量预测的实例分析.结果表明,基于WN-ν-SVM模型的短期预测方法是有效可行的,具有理论意义和实用价值.
Aiming at the characters of multi-dimension,small sample,nonlinearity,randomicity of the time series of product sales,the existing support vector kernel does not accurately approximate any time series curve of the sales. A new wavelet support vector machine(WN-ν-SVM)is proposed on the basis of the combination between wavelet theory and the modified support vector machine. An adaptive and normal mutation particle swarm optimization(ANPSO)algorithm is designed to select the best parameter of WN-ν-SVM model. The application results of vehicle sales prediction case show that the short-term forecasting approach based on the WN-ν-SVM model is more effective and feasible.
作者
彭献永
吴奇
Peng Xianyong;Wu Qi(Emerson Process Management Co.,Ltd.,Shanghai 201206,China;School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《南京师范大学学报（工程技术版）》
CAS
2019年第2期50-58,共9页
Journal of Nanjing Normal University(Engineering and Technology Edition)
基金
国家自然科学基金(61671293)
关键词
自适应变异
正态变异
小波核
支持向量机
短期预测
adaptive mutation
normal mutation
wavelet kernel
support vector machine
short-term forecasting
