【摘要】 为使支持矢量机的参数确定过程更简单高效,提出了一种基于遗传算法的支持矢量机算法。该算法首先确定误差惩罚因子和核参数,利用遗传算法,通过对训练样本自动进化得出对应最优分类超平面,减少了耗时,完成了对确定过程的智能化和参数结果值的最优化;其余两个参数则由支持矢量机算法确定。该算法在Ratsch标准模式库中进行了实验结果的对比,结果表明具有优势,并在一粮食企业的企业资源规划系统资信评估中成功应用,证明该算法具有更好的识别率和更高的性能。更多还原

【Abstract】 To determine the parameters of Support Vector Machine(SVM) simply and efficiently,a SVM based on Genetic Algorithm(GA) was proposed.The tow parameters error castigatory gene and core parameter were determined firstly.By applying GA,the corresponding optimal classification hyper plane was obtained through training the samples automatically,and time consumption was decreased.The intelligent determination process and parameter results optimization were completed.The other two parameters were determined by the SVM algorithm.This algorithm had superiorities in the experiment under the benchmark repository collected by Ratsch,and it was also applied in the Enterprise Resource Planning(ERP) system of a farm corporation for credit rating successfully.Both of them proved that this algorithm had better identification rate and performance.更多还原