摘要
通过分析汽车产品销售时序的特性引入组合预测理论,提出了一种改进的变权重组合预测模型并给出了变权重系数的求取方法。然后针对小样本、多维、多峰、非线性的销售时序特点,采用了基于支持向量机的三种单项预测方法。再通过实例分析显示基于改进变权重组合预测模型的预测精度高于单项预测模型和普通变权重组合预测模型。最后进行了汽车销售时序预测表明基于改进变权重组合预测模型的产品销售预测方法是有效和可行的。
Through the characteristic analysis of car product sale series, a combination forecast theory is introduced, and an improved com- bined forecasting model with variable weights is proposed. And then, the coefficient with variable weights estimation method is put for- ward. Considering the characteristics of multi-dimension, small samples, nonlinearity and multi-apex, three methods of monomial predic- tion based on support vector machine （S.VM） are used in this paper. By analyzing the example, it shows that the improved combined forecasting model with variable weights has higher prediction accuracy than the monomial prediction model. Finally, the results of forecas- ting car sale indicate that the product sale forecasting method based on the improved combined forecasting model with variable weights is effective and feasible.
出处
《计算机技术与发展》
2012年第3期217-221,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(50875046
60934008)
关键词
变权重
组合预测
支持向量机
模型
variable weights
combined forecasting
support vector machine
model
