摘要
针对最小二乘支持向量机(LSSVM)在建模中的重要参数如何选择问题。提出利用具有随机性、遍历性及规律性的混沌优化算法对LSSVM建模过程中的参数进行优化搜索,为了加快对较大搜索空间中的搜索速度,提出变尺度混沌优化算法与遗传算法相结合遗传算法的组合算法对LSSVM中的参数优化。组合算法克服了单一算法存在的早熟、局部收敛及寻优速度慢等问题,把混沌变量种群映射到LSSVM参数取值区间,按照遗传算法训练,同时利用训练集训练LSSVM,最终得到参数优化值。将该方法应用的谷氨酸发酵过程的建模研究,取得了较高建模精度,提高发酵过程资源利用率的同时增加了谷氨酸产量。
To tackle key parameter selection in the process of modeling LSSVM,the paper proposes an optimized search method by using Chaos optimization algorithm which is randomness,ergodicity and regularity.In order to speed up algorithm＇s search velocity in the large space,Chaos optimization algorithm and genetic algorithm are combined to optimize parameter which is used in LSSVM.Premature,local convergence and slow of a single algorithm are gotten over.The chaotic variable populations are mapped to taking value interval of LSSVM parameters,and then train the genetic algorithm,mean while,training LSSVM by training set.Finally,getting the optimal parameters.The proposed method is applied in a soft sensor model for the glutamic acid fermentation process,and the result of simulation shows the effectiveness of the algorithm,the rate of resource and yield in glutamic acid are advanced.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2010年第10期1380-1382,共3页
Computers and Applied Chemistry
基金
国家高技术研究发展计划(863)(2006AA020301).
