【LSSVM分类】基于遗传算法优化LSSVM实现烟叶识别含Matlab源码

matlab科研助手

已于 2022-08-03 16:21:53 修改

阅读量525

点赞数

分类专栏：机器学习组合优化系列matlab完整代码文章标签：大数据

于 2022-07-09 20:44:21 首次发布

本文链接：https://blog.csdn.net/matlab_dingdang/article/details/125698550

版权

机器学习组合优化系列matlab完整代码专栏收录该内容

448 篇文章 90 订阅

订阅专栏

该文提出了一种结合混沌优化和遗传算法的参数优化方法用于最小二乘支持向量机（LSSVM）建模，特别是在烟叶识别任务中。通过优化LSSVM的参数，提高了模型的建模精度。部分代码展示了特征提取和参数优化的过程。仿真结果显示，该方法在烟叶识别中取得了较高的识别率。

摘要由CSDN通过智能技术生成

1 简介

针对最小二乘支持向量机(LSSVM)在建模中的重要参数如何选择问题。提出利用具有随机性、遍历性及规律性的混沌优化算法对LSSVM建模过程中的参数进行优化搜索,为了加快对较大搜索空间中的搜索速度,提出遗传算法对LSSVM中的参数优化。组合算法克服了单一算法存在的早熟、局部收敛及寻优速度慢等问题,把混沌变量种群映射到LSSVM参数取值区间,按照遗传算法训练,同时利用训练集训练LSSVM,最终得到参数优化值。将该方法应用的烟叶识别的建模研究,取得了较高建模精度,。

2 部分代码

function [features,eigvec,eigvals] = AFEm(Xs,kernel, kernel_pars,X,type,nb,eigvec,eigvals) % Automatic Feature Extraction by Nystrom method % % % >> features = AFE(X, kernel, sig2, Xt) % % Description % Using the Nystr m approximation method, the mapping of data to % the feature space can be evaluated explicitly. This gives the % features that one can use for a linear regression or % classification. The decomposition of the mapping to the feature % space relies on the eigenvalue decomposition of the kernel % matrix. The Matlab ('eigs') or Nystr m's ('eign') approximation % using the nb most important eigenvectors/eigenvalues can be % used. The eigenvalue decomposition is not re-calculated if it is % passed as an extra argument. This routine internally calls a cmex file. % % Full syntax % % >> [features, U, lam] = AFE(X, kernel, sig2, Xt) % >> [features, U, lam] = AFE(X, kernel, sig2, Xt, type) % >> [features, U, lam] = AFE(X, kernel, sig2, Xt, type, nb) % >> features = AFE(X, kernel, sig2, Xt, [],[], U, lam) % % Outputs % features : Nt x nb matrix with extracted features % U(*) : N x nb matrix with eigenvectors % lam(*) : nb x 1 vector with eigenvalues % Inputs % X : N x d matrix with input data % kernel : Name of the used kernel (e.g. 'RBF_kernel') % sig2 : parameter of the used kernel % Xt : Data from which the features are extracted % type(*): 'eig'(*), 'eigs' or 'eign' % nb(*) : Number of eigenvalues/eigenvectors used in the eigenvalue decomposition approximation % U(*) : N x nb matrix with eigenvectors % lam(*) : nb x 1 vector with eigenvalues % % See also: % kernel_matrix, RBF_kernel, demo_fixedsize

%FS-LSSVMLab

N = size(X,1); Nc = size(Xs,1);

eval('type;','type=''eig'';'); if ~(strcmp(type,'eig') || strcmp(type,'eigs') || strcmp(type,'eign') ) error('Type needs to be ''eig'', ''eigs'' or ''eign''...'); end

function [model,A,B,C,D] = bay_optimize(model,level, type, nb, bay) % Optimize the posterior probabilities of model (hyper-) parameters with respect to the different levels in Bayesian inference % % One can optimize on the three different inference levels: % % - First level: In the first level one optimizes the support values alpha 's and the bias b. % - Second level: In the second level one optimizes the regularization parameter gam. % - Third level: In the third level one optimizes the kernel % parameter. In the case of the common 'RBF_kernel' the kernel % parameter is the bandwidth sig2. % This routine is only tested with Matlab version 6 using the corresponding optimization toolbox. % % Full syntax % % 1. Outputs on the first level: % % >> [model, alpha, b] = bay_optimize({X,Y,type,gam,sig2,kernel,preprocess}, 1) % >> [model, alpha, b] = bay_optimize(model, 1) % % model : Object oriented representation of the LS-SVM model optimized on the first level % alpha(*) : Support values optimized on the first level of inference % b(*) : Bias term optimized on the first level of inference % % % 2. Outputs on the second level: % % >> [model,gam] = bay_optimize({X,Y,type,gam,sig2,kernel,preprocess}, 2) % >> [model,gam] = bay_optimize(model, 2) % % model : Object oriented representation of the LS-SVM model optimized on the second level of inference % gam(*) : Regularization parameter optimized on the second level of inference % % % 3. Outputs on the third level: % % >> [model, sig2] = bay_optimize({X,Y,type,gam,sig2,kernel,preprocess}, 3) % % model : Object oriented representation of the LS-SVM model optimized on the third level of inference % sig2(*) : Kernel parameter optimized on the third level of inference % % % 4. Inputs using the functional interface % % >> model = bay_optimize({X,Y,type,gam,sig2,kernel,preprocess}, level) % >> model = bay_optimize({X,Y,type,gam,sig2,kernel,preprocess}, level, type) % >> model = bay_optimize({X,Y,type,gam,sig2,kernel,preprocess}, level, type, nb) % % X : N x d matrix with the inputs of the training data % Y : N x 1 vector with the outputs of the training data % type : 'function estimation' ('f') or 'classifier' ('c') % gam : Regularization parameter % sig2 : Kernel parameter (bandwidth in the case of the 'RBF_kernel') % kernel(*) : Kernel type (by default 'RBF_kernel') % preprocess(*) : 'preprocess'(*) or 'original' % level : 1, 2, 3 % type(*) : 'eig', 'svd'(*), 'eigs', 'eign' % nb(*) : Number of eigenvalues/eigenvectors used in the eigenvalue decomposition approximation % % % 5. Inputs using the object oriented interface % % >> model = bay_optimize(model, level) % >> model = bay_optimize(model, level, type) % >> model = bay_optimize(model, level, type, nb) % % model : Object oriented representation of the LS-SVM model % level : 1, 2, 3 % type(*) : 'eig', 'svd'(*), 'eigs', 'eign' % nb(*) : Number of eigenvalues/eigenvectors used in the eigenvalue decomposition approximation % % See also: % bay_lssvm, bay_lssvmARD, bay_modoutClass, bay_errorbar

vers = version; if vers(1)<'6', error(['This routine is only supported currently under MATLAB 6' ... ' and its corresponding optimization toolbox.']); end

if iscell(model), model = initlssvm(model{:}); end

if ~(level==1 | level==2 | level==3), error('level must be 1, 2 or 3.'); end

eval('nb;','nb=model.nb_data;');

if level==1,

eval('type;','type=''train'';'); eval('[F1, F2, F3, C , model] = bay_lssvm(model,1,type,nb,bay);',... '[F1, F2, F3, C , model] = bay_lssvm(model,1,type,nb);'); A = model.alpha B = model.b; elseif level==2,

eval('type;','type=''svd'';'); eval('[model, A,B] = bay_optimize2(model,type,nb,bay); ',... '[model, A,B] = bay_optimize2(model,type,nb);') elseif level==3, % check fminunc resp = which('fminunc'); disp(' '); if isempty(resp), error(' ''fminunc'' not available'); end

eval('type;','type=''svd'';'); %startvalues model = bay_optimize2(model,type,nb); % start value given in model: not fixed 'cause updating % of optimal parameters needs to be possible start_param = model.kernel_pars;

opties=optimset('MaxFunEvals', 250, 'TolFun', .001, 'TolX', .001 ); eval('A = fminunc(@costL3, start_param, opties, model, type, nb);'); model = changelssvm(model,'kernel_pars',abs(A)); [c3,B,model] = bay_lssvm(model,3, type, nb); end

function [model, A,B] = bay_optimize2(model,type,nb,bay)

% check fminunc resp = which('fminunc'); disp(' '); if isempty(resp), error(' ''fminunc'' not available'); end

opties=optimset('TypicalX',model.kernel_pars,'MaxFunEvals', 2000,'GradObj','on','DerivativeCheck', 'off', 'TolFun', .0001, 'TolX', .0001 ); if nargin<4, [c,dc,o, bay] = bay_lssvm(model,2,type,nb); end eval('gam_opt = fminunc(@costL2, abs(model.gam), opties, model, type, nb,bay);'); model = changelssvm(model,'gam',abs(gam_opt)); [D1, D2, D3,B,model] = bay_lssvm(model,2,type, nb, bay); A = model.gam;

function [cost,Dcost] = costL2(lgam, model, type, nb, bay) % model = changelssvm(model,'gam',abs(lgam+1000*eps)); [cost, Dcost] = bay_lssvm(model,2,type, nb, bay);

function cost = costL3(sig2, model, type, nb) % model = changelssvm(model,'kernel_pars',abs(sig2)); cost = bay_lssvm(model,3, type, nb); disp(['sig2 = ' num2str(model.kernel_pars) ' costL3 = ' num2str(cost) ';'])

% eigenvalue decomposition to do.. if nargin<=7, omega = kernel_matrix(Xs, kernel, kernel_pars); if strcmp(type,'eig'), [eigvec,eigvals] = eig(omega+2*eye(size(omega,1))); % + jitter factor eigvals = diag(eigvals); clear omega elseif strcmp(type,'eigs'), eval('nb;','nb=min(size(omega,1),10);'); options.disp = 0; [eigvec,eigvals] = eigs(omega+2*eye(size(omega,1)),nb,'lm',options); clear omega % + jitter factor elseif strcmp(type,'eign'), eval('nb;','nb=min(size(omega,1),10);'); [eigvec,eigvals] = eign(omega+2*eye(size(omega,1)),nb); clear omega % + jitter factor end eigvals = (eigvals-2)/Nc;

peff = eigvals>eps; eigvals = eigvals(peff); eigvec = eigvec(:,peff); clear peff end

if strcmp(kernel,'RBF_kernel') omegaN = sum(X.^2,2)*ones(1,Nc); omegaN = omegaN + ones(N,1)*sum(Xs.^2,2)'; omegaN = omegaN -2*X*Xs'; clear X Xs omegaN = exp(-omegaN./kernel_pars); elseif strcmp(kernel,'lin_kernel') omegaN = X*Xs'; clear X Xs elseif strcmp(kernel,'poly_kernel') omegaN = X*Xs'; clear X Xs omegaN = (omegaN + kernel_pars(1)).^kernel_pars(2); else disp('kernel_type unkown') return; end

%A=sqrt(Nc) ./ sqrt(eigvals); clear eigvals features = omegaN*eigvec; clear omegaN features = repmat((1 ./ sqrt(eigvals))',N,1).*features;

%% 基于遗传算法与最小二乘支持向量机的特征选择的烟叶识别
clc
clear all
close all
% 加载工具箱
addpath('LSSVM工具箱\')
%% 加载数据
data=csvread('最小二乘支持向量机.csv',1,0);
tr_len=36;
[W,D]=size(data);% 数据行和列
% %% 训练数据
X = data(1:tr_len,1:D-1);%输入属性特征数据
Y= data(1:tr_len,D);%标签，即类别
% %% 测试数据
Xt= data(tr_len+1:end,1:D-1);%输入属性特征数据
Yt =data(tr_len+1:end,D);%标签，即类别
%% 归一化
[Xn,inputps] = mapminmax(X');
Xtn= mapminmax('apply',Xt',inputps);
% 转置，符合LSSVM数据格式要求
X=Xn';
Xt=Xtn';
%% LSSVM参数设置
gam=5;
sig2=50;

xlabel('迭代次数')
ylabel('识别率（%）')
legend('平均适应度','最佳适应度')
axis tight
title('GA-LSSVM特征选择')
%% 运用最优特征识别烟叶
L_fold =10;
model = initlssvm(X(:,find(bestchrom==1)),Y,'c',gam,sig2,'RBF_kernel');
model = tunelssvm(model,'simplex','crossvalidatelssvm',{L_fold,'misclass'},'code_OneVsOne');
model = trainlssvm(model);
%% 训练样本
Y_sim = simlssvm(model,X(:,find(bestchrom==1)));
accuracy_1=100*length(find(Y==Y_sim))/length(Y);
figure
plot(1:length(Y),Y,'bo','linewidth',1)
hold on
plot(1:length(Y_sim),Y_sim,'r*','linewidth',1)
xlabel('训练样本','FontSize',12);
ylabel('烟叶类别','FontSize',12);
legend('实际类别','预测类别');
set(gca,'YTick',[0:1:6])
string={'GA-LSSVM识别率';
['accuracy=' num2str(accuracy_1) '%']};
title(string);
grid on
axis tight
%% 测试样本
Yt_sim = simlssvm(model,Xt(:,find(bestchrom==1)));
accuracy_2=100*length(find(Yt==Yt_sim))/length(Yt);
figure
plot(1:length(Yt),Yt,'bo','linewidth',1)
hold on
plot(1:length(Yt_sim),Yt_sim,'r*','linewidth',1)
xlabel('测试样本','FontSize',12);
ylabel('烟叶类别','FontSize',12);
legend('实际类别','预测类别');
set(gca,'YTick',[0:1:6])
string={'GA-LSSVM识别率';
['accuracy=' num2str(accuracy_2) '%']};
title(string);
grid on
axis tight
%% 输出结果
XX=individuals.chrom(end,:);
[m,n]=find(XX==1);
disp(['优化筛选后的输入自变量编号为:' num2str(n)]);