遗传优化核极限学习机的数据分类算法

Genetic optimization kernel-based extreme learning machine data classification algorithm

为了提高核极限学习机(KELM)数据分类的精度,提出了一种结合K折交叉验证(K-CV)与遗传算法(GA)的KELM分类器参数优化方法(GA-KELM),将CV训练所得多个模型的平均精度作为GA的适应度评价函数,为KELM的参数优化提供评价标准,用获得GA优化最优参数的KELM算法进行数据分类。利用UCI中数据集进行仿真,实验结果表明:所提方法在整体性能上优于GA结合支持向量机法(GA-SVM)和GA结合反向传播(GA-BP)算法,具有更高的分类精度。

In order to improve precision of data classification of kernel-based extreme learning machine（KELM）, propose KELM classification parameter optimization method,GA-KELM,which combines K-fold cross-validation （K-CV） and genetic algorithms（GA）,the average precision of multiple models of resulting of CV training as GA fitness evaluation function value,provide evaluation criteria for parameter optimization of KELM ,and then the KELM algorithm is used to get the optimization parameters of GA for data classification. Using UCI dataset for simulation,results show that the proposed method is superior to GA-SVM and GA-BP algorithm on the overall performance,with a higher classification precision.