CAJ | 학술논문

粒计算是信息处理邻域中新的概念和计算方法，但是，传统粒子支持向量机算法存在着映射前后的数据分布不一致的问题，同时，由于使用粒子中心替代粒子从而导致精度下降。为此，提出基于映射后的混合粒子支持向量机算法，首先，利用mercer核函数将数据映射到高维空间，粒化计算后，找出含有更多分类信息的混合粒子，提取后作为输入集合对超平面进行训练，利用几何分析调整最优超平面，并采用基于QPSO算法对关键参数进行最优求解，从而提高算法的精度。实验表明该算法比传统粒子支持向量机算法正确率高10％，说明改进的粒化支持向量机算法提升效果明显。
립계산시신식처리린역중신적개념화계산방법，단시，전통입자지지향량궤산법존재착영사전후적수거분포불일치적문제，동시，유우사용입자중심체대입자종이도치정도하강。위차，제출기우영사후적혼합입자지지향량궤산법，수선，이용mercer핵함수장수거영사도고유공간，립화계산후，조출함유경다분류신식적혼합입자，제취후작위수입집합대초평면진행훈련，이용궤하분석조정최우초평면，병채용기우QPSO산법대관건삼수진행최우구해，종이제고산법적정도。실험표명해산법비전통입자지지향량궤산법정학솔고10％，설명개진적립화지지향량궤산법제승효과명현。
Granular computing is a new concept and computing paradigm in the domain of information processing. However, there is an inconsistent problem with traditional granular support vector machine since granulation computa-tion is usually done after mapping into high dimensional space and poor performance is caused by replacing whole granular with center for traditional GSVM.This paper presents a novel granular support vector machine algorithm based on mixed granules to solve the existing problems.The original data will be mapped into high dimensional space by mercer kernel.Then the data are divided into some granules.Mixed granules are extracted and used to train SVM and hyperplane is further corrected using geometric analyzing.And optimal parameters are obtained by using QPSO. The experiment results show that the accuracy rate of the algorithm is improved by 10% compared to traditional GSVM.