结合概率和等价类的双系数支持向量机

Double coefficients support vector machine with probability and equivalence class

支持向量机(SVM)的有效性依赖于对数据信息获取的准确性。针对传统SVM模型获取数据信息单一导致分类精度不高、泛化能力不强的问题,结合概率分布特性和等价类关系,提出了一种双系数控制分类的新模型。该模型优化了SVM,以双系数方式改进传统参数,为每一个样本同时赋予概率值和等价类系数,充分挖掘数据信息内在规律和联系。实验结果证明:该模型能有效利用数据信息,与SVM、FSVM和RSVM相比有较高的测试精度,能有效提高分类能力,具有较高鲁棒性。

The effectiveness of a Support Vector Machine （SVM） depends on the accuracy of the acquired data information. Considering the low prediction accuracy and poor generalization capacity of SVM caused by simply mining data, a new model was proposed. Combining probability distribution and equivalence class information among data, this model optimized the traditional SVM by adopting double coefficients to promote the capability of acquiring information. Each instance will be assigned two coefficients of probability value and equivalence class. The comparison with SVM, Rough Support Vector Machine （RSVM） and Fuzzy Support Vector Machine （FSVM） illustrates the new model can not only utilize information effectively but also assure a remarkable prediction accuracy and classification ability, and be more robust.