摘要
基于统计学习的预测方法在一定程度上依赖于大量数据的基础假设,但在实际应用中时间序列样本往往是有限的,论文提出了一种基于小波的支持向量机预测模型(WSVMFM)。通过小波框架的平移所生成的平方可积空间中的一组完备的基可以构造为支持向量机(SVM)的核函数,而采用该核函数的 SVM(WSVM)可以逼近平方可积空间中的任意函数,从而提高学习和预测效率。将该预测模型应用于基于多智能代理的电子商务交易模型中可较好地解决交易状态表示及预测等问题。
The forecasting method based on statistic learning depends on the basic hypothesis of mass data, but in the really application, the time serial sample always is limited, this paper puts forward a forecasting model via support vector machines based wavelets（WSVMFM）. A set of maturity radix in square accumulated space generated by smooth shifting the frame of wavelets could construct the kernel function of SVM, and that the WSVM can approach any function in square accumulated space. Sequentially improve the efficiency of learning and forecasting. This paper also validated the model＇s correctness and it＇s validity by experiment and simulation.
出处
《计算机科学》
CSCD
北大核心
2006年第3期202-204,共3页
Computer Science
基金
重庆市重点科技攻关资助项目(7220-B-12)
重庆市自然科学基金(CSTC
2004BB2167)
关键词
小波框架
支持向量机
预测模型
电子商务交易模型
Frame of wavelets, Support vector machine, Forecasting model, Electronic business tradeoff model
