【摘要】 提出一种区分不同类别误分代价的支持向量机模型,建立了拉格朗日方程,利用KKT条件推导出其对偶形式.该模型能够让用户标记不同类别样本的重要性,并根据重要性程度设定不同的误分代价,使重要类别的样本尽量减少被误分的可能性,以满足特定任务的需求.实验结果表明,该模型可以很好地完成特定任务的分类.更多还原

【Abstract】 A support vector machine model that distinguishes different samples category’s penalty coefficients is proposed.The Lagrange equation is built and its dual form of modal is deduced by KKT condition.This model could let users mark the importance of different samples,and give relevant penalty coefficients according to different importance.This situation can decrease the possibility of being misclassified in important samples.The result of experimentation shows that this model can classify customizing task well.更多还原