论文部分内容阅读
针对粒子群(PSO)算法优化支持向量机(SVM)参数存在容易陷入局部最优的问题,通过引入新的动态惯性权重、全局邻域搜索、收缩因子和遗传算法中的变异操作,提出了一种基于改进粒子群(IPSO)算法优化SVM参数(IPSO-SVM)的改进型分类器。采用UCI机器学习库中的公共数据集Iris、Wine和seeds来测试其分类效果,结果表明IPSO-SVM分类器在分类准确率和分类时间上优于GS-SVM、AFSA-SVM、GA-SVM和PSO-SVM分类器。最后,将IPSO-SVM分类器应用于Sallen-Key带通滤波器、四运放双二次高通滤波器及非线性整流电路的故障诊断中,结果表明IPSO-SVM分类器具有较强的全局收敛能力和较快的收敛速度。
In order to solve the problem of Particle Swarm Optimization (PSO) based optimization support vector machine (SVM), the parameters of SVM easily fall into the local optimum. By introducing new dynamic inertia weight, global neighborhood search, shrinkage factor and mutation operation in genetic algorithm, Improved classifier based on improved particle swarm optimization (IPSO-SVM) algorithm. The classification results of ISOs, Wine and seeds were tested using public data sets Iris, Wine and seeds in UCI machine learning library. The results show that IPSO-SVM classifier is better than GS-SVM, AFSA-SVM and GA-SVM in classification accuracy and classification time PSO-SVM classifier. Finally, the IPSO-SVM classifier is applied to fault diagnosis of Sallen-Key bandpass filter, quadrature amplifier quadratic quadratic high-pass filter and nonlinear rectifier circuit. The results show that IPSO-SVM classifier has strong global convergence Ability and faster convergence rate.