摘要
结合核主成分分析的主因子提取和支持向量机的分类机理,提出了一种组合建模算法。应用核主成分分析过程作为预处理器,可以把共线性的多元变量糅合为几个主因子,但基本不损失有效信息。然后进行基于支持向量机的分类建模和预测。这样不仅可以防止共线性多元变量对模型的负面影响,还可以降低数据维数,减少支持向量机分类过程中的复杂度和运算量。最后用实验进行评估所得到的训练模型,实例说明了所提方法的有效性。
A novel hybrid algorithm based on principal components and classification principles of support vector machines is presented. The principal components analysis is applied as preprocessor so that the redundant attributes are deleted, and the efficient information is remain. Then, classification modeling and forecasting test based on SVM are put in practice. By this method, the negative influence of multi-variable＇s redundant attributes is greatly reduced, and the complexity in the process of SVM classification is highly decreased. The occupied memory is cut down. Finally, the experimental result show the effectiveness of the suggested hybrid method.
出处
《计算机工程与设计》
CSCD
北大核心
2006年第8期1385-1388,共4页
Computer Engineering and Design
关键词
核主成分分析
支持向量机算法
多元共线性
核函数
分类算法
机器学习
kernel principal components analysis
support vector machine
redundant attributes
kernel function
classification algorithm
machine learning
