摘要
针对最小二乘支持向量机(LSSVM)处理大数据集时确定最优模型参数耗时长、占内存大的问题,提出了一种基于人工免疫算法的参数寻优方法。通过分析LSSVM模型参数对分类准确率的影响发现,存在多种参数组合,使得分类准确率相同;当其中一个参数固定,另外一个参数在某些范围内变化取值时,它们的组合并不影响分类的准确率。将LSSVM模型参数作为抗体的基因设计了抗体的编码方案,利用人工免疫算法对LSSVM参数优化搜索。仿真结果表明,与使用交叉验证和网格搜索方法相比,提出的LSSVM参数优化算法在不降低分类准确率的前提下,寻优效率大大提高。
To reduce training time of least squares support vector machine （LSSVM） on large datasets,this paper presented a novel algorithm for selecting LSSVM optimal parameters which was based on principle of artificial immune. By analyzing LSSVM parameters on the classification accuracy rate,it was found that the existence of many parameters combinations which made the same classification accuracy rate.What’s more,once one of the parameters fixed and the other changed in a certain range, the combinations of them didn’t affect the classification accuracy rate.By making LSSVM parameters as antibody genes,this paper designed reasonable coding scheme for these antibodies.Then employed artificial immune algorithm to search the optimal parameters of LSSVM.Simulation results show that the proposed algorithm greatly enhance parameters optimizing efficiency while keeping the same classification accuracy rate comparing with the method of cross-validation or grid- search.
出处
《计算机应用研究》
CSCD
北大核心
2010年第5期1702-1704,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60673153
60970105)
山东省自然科学基金资助项目(Y2007G22)
关键词
人工免疫算法
最小二乘支持向量机
参数优化
artificial immune algorithm
least squares support vector machines
parameters optimization
