基于SVM的多分类器构造算法的研究

On Research of Algorithms about Structuring Multi-Classifier Based on SVM

在对传统的多类分类算法研究的基础上,针对基于二值分类器的多分类器构造算法存在的预测精度低、训练时间长的缺点,提出了一种基于SVM的组合回归机构造多类分类器的算法。该算法解决了二值分类器方法中存在的信息丢失问题,同时避免了由于参数调整而造成的计算代价过大的问题。实验结果表明:新的SVM多分类算法大大降低了计算代价,提高了运行效率和预测的精度,减少了运行时间。

By researching the foundation of the traditional multi - classification algorithms, the algorithm about structuring multi- classifier based on SVM is proposed by combining regression machines in this paper. The algorithm can resolve the problem of how to prevent the loss of information which occurs in the usual bi - classifier algorithms, and can avoid the problem of calculated price over great because of adjusting parameter. The results show that the new algorithm decreases the calculated price and the circulated time consumedly so that it enhances the running efficiency and the estimating accuracy.