摘要
为解决支持向量机中核函数的参数优化选择问题,在对粒子群算法中惯性权重和加速因子非线性化的基础上,提出一种动态非线性策略的粒子群优化算法。算法的核心是通过调整和融合惯性权重ω和加速因子c1,c2选择策略,有效控制算法的全局寻优与局部寻优能力,限定粒子的搜索范围。采用单模态和多模态标准测试函数检验策略对算法的影响,并将该算法应用于标准支持向量机非线性测试函数的拟合问题中,最后应用优化后的支持向量机解决航空发动机振动监控问题。仿真结果表明,改进后算法能有效提高最优解精度,加快收敛速度,实现支持向量机参数的择优选取,具有良好的工程应用价值。
In order to resolve the parameter optimizing selection problem of kernel function in support vector machine（SVM）,based on the foundation of non-linearization of inertia weight and accelerated constant in particle swarm optimization（PSO）,a particle swarm optimization was proposed based on dynamic nonlinear strategy（DNLSPSO）.The core of algorithm was to control the ability of globel and local optimization and to limit the search range of particles by adjusting and fusing the selection strategy of inertia weight ω and accelerated constant c1,c2.Then the influence of the strategy on the algorithm was verified by using single mode and multi mode standard test functions,and the algorithm was applied to the regression problem of standard support vector machine nonlinear test function.Finally,the problem of aeroengine vibration monitoring was solved by the optimized support vector machine.Simulation result shows that the improved algorithm can increase the precision of optimization solution efficiently,quicken the speed of convergence,and achieve the fine parameter selection of support vector machine;and the value of engineering application is well.
出处
《计算机仿真》
CSCD
北大核心
2012年第10期122-126,278,共6页
Computer Simulation
基金
航空科学基金(20100818017)
陕西省电子信息系统综合集成重点实验室基金(201113y12)
关键词
支持向量机
粒子群优化
非线性策略
参数优化
状态监控
Support vector machine（SVM）
Particle swarm optimization（PSO）
Nonlinear strategy
Parameter Optimization
Condition monitoring
