遗传算法(GA)优化核极限学习机回归，GA-KELM回归预测，多变量输入模型 评价指标包括:R2、MAE、MSE、RMSE和

%_________________________________________________________________________% %基于GA算法优化的KELM回归问题 %_________________________________________________________________________% % clc;clear;close all % %% 读取数据 % %res = xlsread('数据集.xlsx'); % res1 = importdata('待预测data.txt'); % res = datarun_fun(res1,5); function [TestOutT,T_test,net,R2] = KELM_FUN(res,num_size,dataran) %% 数据分析 % num_size = 0.98; % 训练集占数据集比例 outdim = 1; % 最后一列为输出 num_samples = size(res, 1); % 样本个数 num_train_s = round(num_size * num_samples); % 训练集样本个数 AA1 = res(1:num_train_s,:); AA2 = res(num_train_s+1:end-1,:); AA3 = res(end,:); clear res; if dataran == 1 AA1 = AA1(randperm(size(AA1, 1)), :); % 打乱数据集（不希望打乱时，注释该行） AA2 = AA2(randperm(size(AA2, 1)), :); % 打乱数据集（不希望打乱时，注释该行） end res=[AA1;AA2;AA3]; f_ = size(res, 2) - outdim; % 输入特征维度 %% 划分训练集和测试集 P_train = res(1: num_train_s, 1: f_)'; T_train = res(1: num_train_s, f_ + 1: end)'; M = size(P_train, 2); P_test = res(num_train_s + 1: end, 1: f_)'; T_test = res(num_train_s + 1: end, f_ + 1: end)'; N = size(P_test, 2); %% 归一化 % 训练集 [Pn_train,inputps] = mapminmax(P_train,-1,1); Pn_test = mapminmax('apply',P_test,inputps); % 测试集 [Tn_train,outputps] = mapminmax(T_train,-1,1); Tn_test = mapminmax('apply',T_test,outputps); %% 参数设置 pop=20; %种群数量 Max_time=20; % 设定最大迭代次数 dim = 2;% 维度为2，即优化两个参数，正则化系数 C 和核函数参数 S lb = [1;1];%下边界 ub = [20;20];%上边界 fobj = @(x) KELM_1fun(x,Pn_train,Tn_train); [Best_pos,Best_score,curve]=KELM_GA(pop,Max_time,lb,ub,dim,fobj); %开始优化 % figure % plot(curve,'linewidth',1.5); % grid on; % xlabel('迭代次数') % ylabel('适应度值') % title('收敛曲线') %% 获取最优正则化系数 C 和核函数参数 S Regularization_coefficient = Best_pos(1); Kernel_para = Best_pos(2); Kernel_type = 'rbf'; %% 训练 [TrainOutT,OutputWeight] = KELM_kelmTrain(Pn_train,Tn_train,Regularization_coefficient,Kernel_type,Kernel_para); %% 训练集预测 InputWeight = OutputWeight; [TestOutT] = KELM_kelmPredict(Pn_train,InputWeight,Kernel_type,Kernel_para,Pn_test); net = [InputWeight;Kernel_para]; %% 训练集正确率 TrainOutT = mapminmax('reverse',TrainOutT,outputps); errorTrain = TrainOutT - T_train; MSEErrorTrain = mse(errorTrain); %% 测试集正确率 TestOutT = mapminmax('reverse',TestOutT,outputps); errorTest = TestOutT - T_test; MSEErrorTest = mse(errorTest); N1=length(T_test); % R2 = (N1*sum(TestOutT.*T_test)-sum(TestOutT)*sum(T_test))^2/((N1*sum((TestOutT).^2)-(sum(TestOutT))^2)*(N1*sum((T_test).^2)-(sum(T_test))^2)); testnum=size(TestOutT,2); zdre=0; for io=1:testnum-1 AA1=TestOutT(:,io); AA2=TestOutT(:,io+1); BB1=T_test(:,io); BB2=T_test(:,io+1); % if or(and(AA1>AA2,BB1>BB2),and(AA1<AA2,BB1<BB2)) if abs(AA1-BB1)<0.5; zdre=zdre+1; end end R2 = zdre/testnum; end