摘要
目前回归函数中普遍存在的泛化能力得不到保证的缺点,结合统计学习理论的研究成果,建立了基于最小一乘准则的最优回归模型(LaOR模型)。与以往回归模型相比较,新模型综合考虑了回归误差和置信范围,可望有效地降低回归模型的期望风险。上海市将LaOR应用到物流需求的短期预测中,取得了可以接受的预测效果。
Aimming at the disadvantages of weak generalization ability that exists in most of the current regression functions, combining with the research achievement of statistic learning theory, the paper proposes the optimal regress model based on least - absolute criteria, or LaOR model. Compared with other regress models, LaOR model has taken regress error and confidence interval into account synthetically. LaOR model can reduce the expected risk of regress model effectively. The logistics demand short - term forecasting of Shanghai is used as an example to examine the validity of the LaOR model.
出处
《商业研究》
北大核心
2007年第9期32-35,共4页
Commercial Research
基金
中国博士后科学基金资助
项目编号:2005038042
广东省社会发展领域科技计划项目资助
项目编号:63121
关键词
最小一乘准则
统计学习理论
多元回归
物流预测
least - absolute criteria
statistic learning theory
multiple regression
logistics forecasting
