【数据分析】基于粒子群优化的模糊专家系统附matlab代码

✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，matlab项目合作可私信。

🍎个人主页：Matlab科研工作室

🍊个人信条：格物致知。

⛄ 内容介绍

 PSO 神经网络分类 % 所以，网上没有合适的进化分类 Matlab 代码， % 决定用 PSO 做一个。% 此代码获取用于分类的数据输入，其中包含数据和 % 标签并将其存储到“netdata”中。数据由 6% 类的 300 个样本组成，其中包括 40 个特征。您可以提取您的特征并 % 标记它，因为它是一个监督模型。这些特征是从小物体图像中提取的 SURF % 特征。现在，System 是 PSO % 和典型的浅层神经网络的组合。神经网络本身构成了系统的 % 初始结构或主体，但 PSO 有责任对训练中的 % 神经元进行加权，进化算法的真正力量在这里体现。最后与SVM、KNN的结果对比，和 TREE 分类 % 算法作为混淆矩阵和最终识别准确率。% 有三个重要参数 'NH'（隐藏层数），% 'SwarmSize' 和 'MaxIteration' 会显着影响系统的性能。因此，为了获得理想的结果，您应该根据您的数据使用这些参数来玩 %。这里唯一的缺点是 % 标签是手动完成的，你可以很容易地自己修复它，但 % 是在训练的主要阶段之后。此代码可以扩展为使用其他进化算法（例如 GA 或 DE）进行训练。如果您发现 % 有任何问题，请按以下方式与我联系：这会显着影响您 % 系统的性能。因此，为了获得理想的结果，您应该根据您的数据使用这些参数来玩 %。这里唯一的缺点是 % 标签是手动完成的，你可以很容易地自己修复它，但 % 是在训练的主要阶段之后。此代码可以扩展为使用其他进化算法（例如 GA 或 DE）进行训练。如果您发现 % 有任何问题，请按以下方式与我联系：这会显着影响您 % 系统的性能。因此，为了获得理想的结果，您应该根据您的数据使用这些参数来玩 %。这里唯一的缺点是 % 标签是手动完成的，你可以很容易地自己修复它，但 % 是在训练的主要阶段之后。此代码可以扩展为使用其他进化算法（例如 GA 或 DE）进行训练。

⛄ 部分代码

%% PSO Neural Network Classification

% So, there was no proper evolutionary classification Matlab code in the web,

% Which decided to make one with PSO.

% This code gets data input for classification which contains data and 

% labels and stores it into 'netdata'. data consists of 300 samples for 6 

% classes which includes 40 features. You can extract your features and

% label it as it is a supervised model. These features are extracted SURF

% features out of small objects images. Now, System is combination of PSO

% and typical shallow neural network. Neural network itself makes the

% initial structure or body of the system but PSO has duty of weighting the

% neurons in training, which true power of evolutionary algorithms present

% here. Finally, result compared with SVM, KNN, and TREE classification 

% algorithms as confusion matrix and final recognition accuracy.

% There are three important parameters of 'NH' (number of hidden layers),

% 'SwarmSize' and 'MaxIteration' which effect the performance of thee

% system significantly. So, in order to get desired result, you should play

% with these parameters based on your data. The only drawback here is that

% labeling is done manually, which you can fix it yourself easily, but it

% is after main stage of training. This code could be expanded to be

% trained with other evolutionary algorithms such as GA or DE. If you find

% any problem, please contact me as below:

% Email: mosavi.a.i.buali@gmail.com

% Author: Seyed Muhammad Hossein Mousavi

% Also this code is part of the following project, so please cite below 

% after using the code: 

% Mousavi, Seyed Muhammad Hossein, et al. "A PSO fuzzy-expert system: As an

% assistant for specifying the acceptance by NOET measures, at PH. D 

% level." 2017 Artificial Intelligence and Signal Processing Conference

% (AISP). IEEE, 2017.

% Thank you for citing the paper and enjoy the code (hope it help you (Be happy :)

%%

warning('off');

% Data Loading

clear;

netdata=load('fortest2.mat');

netdata=netdata.FinalReady;

% Data and Label

network=netdata(:,1:end-1);

netlbl=netdata(:,end);

% Var Change

inputs = network;

targets = netlbl;

% Dim Size

InputNum = size(inputs,2);

OutputNum = size(targets,2);

pr = [-1 1];

PR = repmat(pr,InputNum,1);

% NN Structure (log-sigmoid transfer function)

NH=5;     % Number of Hidden Layers (more better)

Network1 = newff(PR,[NH OutputNum],{'tansig' 'tansig'});

% Train with PSO on Networks Weights

Network1 = TrainPSO(Network1,inputs,targets);

view(Network1)

% Generating Outputs from Our PSO + NN Network Model

outputs = Network1(inputs');

outputs=outputs';

% Size

sizenet=size(network);

sizenet=sizenet(1,1);

% Outputs Error

MSE=mse(outputs);

% Bias Output for Confusion Matrix

outputs=outputs-(MSE*0.1)/2;

% Detecting Mislabeled Data

for i=1 : 50

            if outputs(i) <= 0.9

               out(i)=0;

        elseif outputs(i) >= 0.9

               out(i)=1;            end;end;

for i=51 : 100

            if outputs(i) <= 0.9

               out(i)=0;

        elseif outputs(i) >= 0.9

               out(i)=2;            end;end;

for i=101 : 150

            if outputs(i) <= 0.9

               out(i)=0;

        elseif outputs(i) >= 0.9

               out(i)=3;            end;end;

for i=151 : 200

            if outputs(i) <= 0.9

               out(i)=0;

        elseif outputs(i) >= 0.9

               out(i)=4;            end;end;

for i=201 : 250

            if outputs(i) <= 0.9

               out(i)=0;

        elseif outputs(i) >= 0.9

               out(i)=5;            end;end;

for i=251 : 300

            if outputs(i) <= 0.9

               out(i)=0;

        elseif outputs(i) >= 0.9

               out(i)=6;            end;end;

       out1=single(out');

% PSO Final Accuracy

       psomse=mse(out1,targets);

       MSEError=abs(mse(targets)-mse(out1));

       cnt=0;

       for i=1:sizenet

           if out1(i)~= targets(i)

               cnt=cnt+1;

           end;

       end;

      fin=cnt*100/ sizenet;

      psoacc=(100-fin)-psomse;

%

%% KNN for Comparison 

lblknn=netdata(:,end);

dataknn=netdata(:,1:end-1);

Mdl = fitcknn(dataknn,lblknn,'NumNeighbors',8,'Standardize',1);

rng(1); % For reproducibility

knndat = crossval(Mdl);

classError = kfoldLoss(knndat);

% Predict the labels of the training data.

predictedknn = resubPredict(Mdl);

ctknn=0;

for i = 1 : sizenet(1,1)

if lblknn(i) ~= predictedknn(i)

    ctknn=ctknn+1;

end;

end;

finknn=ctknn*100/ sizenet;

KNN=(100-finknn)-classError;

%

%% SVM for Comparison

tsvm = templateSVM('KernelFunction','polynomial');

svmclass = fitcecoc(dataknn,lblknn,'Learners',tsvm);

svmerror = resubLoss(svmclass);

CVMdl = crossval(svmclass);

genError = kfoldLoss(CVMdl);

% Predict the labels of the training data.

predictedsvm = resubPredict(svmclass);

ct=0;

for i = 1 : sizenet(1,1)

if lblknn(i) ~= predictedsvm(i)

    ct=ct+1;

end;

end;

% Compute Accuracy

finsvm=ct*100/ sizenet;

SVMAccuracy=(100-finsvm);

%% Tree for Comparison

Mdl2 = fitctree(dataknn,lblknn);

rng(1); % For reproducibility

treedat = crossval(Mdl2);

classErrortree = kfoldLoss(treedat);

% Predict the labels of the training data.

predictedtree = resubPredict(Mdl2);

cttree=0;

for i = 1 : sizenet(1,1)

if lblknn(i) ~= predictedtree(i)

    cttree=cttree+1;

end;

end;

fintree=cttree*100/ sizenet;

TREE=(100-fintree)-classErrortree;

%% Plots and Results

% Confusion Matrix

figure

% set(gcf, 'Position',  [50, 100, 1300, 300])

subplot(2,2,1)

cmsvm = confusionchart(lblknn,predictedsvm);

cmsvm.Title = (['SVM Classification =  ' num2str(SVMAccuracy) '%']);

subplot(2,2,2)

cmknn = confusionchart(lblknn,predictedknn);

cmknn.Title = (['KNN Classification =  ' num2str(KNN) '%']);

subplot(2,2,3)

cmtree = confusionchart(lblknn,predictedtree);

cmtree.Title = (['Tree Classification =  ' num2str(TREE) '%']);

subplot(2,2,4)

cmpso = confusionchart(out1,targets);

cmpso.Title = (['PSO-NN Classification =  ' num2str(psoacc) '%']);

% Regression

figure

set(gcf, 'Position',  [50, 150, 450, 350])

[population2,gof] = fit(targets,out1,'poly4');

plot(targets,out1,'o',...

    'LineWidth',3,...

    'MarkerSize',5,...

    'Color',[0.3,0.9,0.2]);

    title(['PSO - R =  ' num2str(1-gof.rmse)]);

    xlabel('Train Target');

    ylabel('Train Output');   

hold on

plot(population2,'b-','predobs');

    xlabel(' Target');

    ylabel(' Output');   

hold off

% ACC and Metrics Results

fprintf('The SVM Accuracy is = %0.4f.\n',SVMAccuracy)

fprintf('The KNN Accuracy is = %0.4f.\n',KNN)

fprintf('The Tree Accuracy is = %0.4f.\n',TREE)

fprintf('The PSO Accuracy is = %0.4f.\n',psoacc)

fprintf('PSO MSE is = %0.4f.\n',MSEError)

fprintf('PSO RMSE is = %0.4f.\n',sqrt(MSEError))

fprintf('PSO MAE is = %0.4f.\n',mae(targets,out1))

⛄ 运行结果

⛄ 参考文献

⛄ Matlab代码关注

❤️部分理论引用网络文献，若有侵权联系博主删除

❤️ 关注我领取海量matlab电子书和数学建模资料