基于改进磷虾群算法的SVDD参数优化

发布时间：2018-06-09 05:07

本文选题：支持向量数据描述 + 改进磷虾群算法　；参考：《计算机工程与应用》2017年22期

【摘要】：支持向量数据描述(SVDD)是构造单类数据描述的分类算法,惩罚参数C和核参数σ作为影响SVDD分类效果的关键,其合理选取一直是个难点。针对这一问题,提出了一种基于改进磷虾群算法的SVDD参数优化算法(IKH-SVDD)。依据仿真实验,分析参数C和σ对描述边界的影响;引入磷虾群算法并分析其优劣,通过在随机扩散行为中定义扰动因子,增强算法的全局搜索能力;将一种新的精英选择和保留策略引入迭代过程,提高算法的收敛精度;将改进的磷虾群算法引入SVDD参数优化过程,构建了IKH-SVDD参数优化模型。基于UCI标准数据库进行实验并与其他几种参数优化算法进行比较,结果表明了IKH-SVDD算法具有更高的分类准确性。
[Abstract]:Support vector data description (SVDDs) is a classification algorithm for constructing single-class data description. The penalty parameter C and kernel parameter 蟽 are the key to affect the classification effect of SVDD. The reasonable selection of SVDD classification is always a difficulty. In order to solve this problem, a SVDD parameter optimization algorithm based on improved krill swarm algorithm is proposed. According to the simulation experiment, the influence of parameters C and 蟽 on describing the boundary is analyzed, and the algorithm of krill swarm is introduced and its merits and demerits are analyzed. The global searching ability of the algorithm is enhanced by defining the disturbance factor in the random diffusion behavior. A new elite selection and retention strategy is introduced into the iterative process to improve the convergence accuracy of the algorithm, and the improved krill swarm algorithm is introduced into the optimization process of SVDD parameters, and the IKH-SVDD parameter optimization model is constructed. Experiments based on UCI standard database are carried out and compared with other parameter optimization algorithms. The results show that IKH-SVDD algorithm has higher classification accuracy.
【作者单位】：空军工程大学防空反导学院;
【基金】：国家自然科学基金(No.61273275)
【分类号】：TP18

【相似文献】