摘要
针对不同类别样本数差异和不同误分代价的分类问题,提出了一种基于最小二乘加权支持向量机的分类预测方法。在最小二乘加权支持向量机的基础上,考虑不同类别样本数差异和不同误分代价,提出了新的最小二乘加权支持向量机分类模型,构造了新的最优分类函数。将该模型应用于个人信用预测实验,与已有方法的对比实验结果表明,提出的模型在解决不同类别样本数差异和不同误分代价的个人信用预测问题时,有效地降低了总误分代价,提高了个人信用预测精确度。
A new classifying prediction method based on least squares support vector machines is proposed for the problem of classification with uneven class sizes and different costs of misclassification. Least squares support vector machines are introduced firstly. When uneven class sizes and different costs of misclassification are considered, a new classification model of least squares support vector machines is proposed, and a new optimal classifying function is formulated. Then the proposed model is used to predict the personal credit. Experimental results on both sample learning and the prediction show that total costs of misclassification decrease using the proposed method compared with existing methods, and the precision of the prediction for the personal credit is improved effectively.
出处
《运筹与管理》
CSCD
2008年第4期89-95,共7页
Operations Research and Management Science
基金
国家自然科学基金资助项目(70471037)
陕西省自然科学基金资助项目(2004G02)
关键词
最小二乘支持向量机
加权支持向量机
类别差异
误分代价
个人信用预测
least squares support vector machines
weighted support vector machines
uneven class size
cost ofmisclassification
prediction of the personal credit
