%
% Copyright (c) 2015, Yarpiz (www.yarpiz.com)
% All rights reserved. Please read the "license.txt" for license terms.
%
% Project Code: YPFZ101
% Project Title: Nonlinear Regression using ANFIS
% Publisher: Yarpiz (www.yarpiz.com)
%
% Developer: S. Mostapha Kalami Heris (Member of Yarpiz Team)
%
% Contact Info: sm.kalami@gmail.com, info@yarpiz.com
%
clc;
clear;
close all;
%% Create Time-Series Data
data = load('engine_data');
% About Engine Dataset
% Inputs:
% 1. Fuel Rate
% 2. Speed
% Targets:
% 1. Torque
% 2. Nitrous Oxide Emissions
Inputs = data.Inputs';
Targets = data.Targets';
nData = size(Inputs,1);
Targets = Targets(:,2); % Select 1st Output to Model
%% Shuffling Data
PERM = randperm(nData); % Permutation to Shuffle Data
pTrain=0.85;
nTrainData=round(pTrain*nData);
TrainInd=PERM(1:nTrainData);
TrainInputs=Inputs(TrainInd,:);
TrainTargets=Targets(TrainInd,:);
pTest=1-pTrain;
nTestData=nData-nTrainData;
TestInd=PERM(nTrainData+1:end);
TestInputs=Inputs(TestInd,:);
TestTargets=Targets(TestInd,:);
%% Selection of FIS Generation Method
Option{1}='Grid Partitioning (genfis1)';
Option{2}='Subtractive Clustering (genfis2)';
Option{3}='FCM (genfis3)';
ANSWER=questdlg('Select FIS Generation Approach:',...
'Select GENFIS',...
Option{1},Option{2},Option{3},...
Option{3});
pause(0.01);
%% Setting the Parameters of FIS Generation Methods
switch ANSWER
case Option{1}
Prompt={'Number of MFs','Input MF Type:','Output MF Type:'};
Title='Enter genfis1 parameters';
DefaultValues={'5', 'gaussmf', 'linear'};
PARAMS=inputdlg(Prompt,Title,1,DefaultValues);
pause(0.01);
nMFs=str2num(PARAMS{1}); %#ok
InputMF=PARAMS{2};
OutputMF=PARAMS{3};
fis=genfis1([TrainInputs TrainTargets],nMFs,InputMF,OutputMF);
case Option{2}
Prompt={'Influence Radius:'};
Title='Enter genfis2 parameters';
DefaultValues={'0.3'};
PARAMS=inputdlg(Prompt,Title,1,DefaultValues);
pause(0.01);
Radius=str2num(PARAMS{1}); %#ok
fis=genfis2(TrainInputs,TrainTargets,Radius);
case Option{3}
Prompt={'Number fo Clusters:',...
'Partition Matrix Exponent:',...
'Maximum Number of Iterations:',...
'Minimum Improvemnet:'};
Title='Enter genfis3 parameters';
DefaultValues={'15', '2', '200', '1e-5'};
PARAMS=inputdlg(Prompt,Title,1,DefaultValues);
pause(0.01);
nCluster=str2num(PARAMS{1}); %#ok
Exponent=str2num(PARAMS{2}); %#ok
MaxIt=str2num(PARAMS{3}); %#ok
MinImprovment=str2num(PARAMS{4}); %#ok
DisplayInfo=1;
FCMOptions=[Exponent MaxIt MinImprovment DisplayInfo];
fis=genfis3(TrainInputs,TrainTargets,'sugeno',nCluster,FCMOptions);
end
%% Training ANFIS Structure
Prompt={'Maximum Number of Epochs:',...
'Error Goal:',...
'Initial Step Size:',...
'Step Size Decrease Rate:',...
'Step Size Increase Rate:'};
Title='Enter genfis3 parameters';
DefaultValues={'200', '0', '0.01', '0.9', '1.1'};
PARAMS=inputdlg(Prompt,Title,1,DefaultValues);
pause(0.01);
MaxEpoch=str2num(PARAMS{1}); %#ok
ErrorGoal=str2num(PARAMS{2}); %#ok
InitialStepSize=str2num(PARAMS{3}); %#ok
StepSizeDecreaseRate=str2num(PARAMS{4}); %#ok
StepSizeIncreaseRate=str2num(PARAMS{5}); %#ok
TrainOptions=[MaxEpoch ...
ErrorGoal ...
InitialStepSize ...
StepSizeDecreaseRate ...
StepSizeIncreaseRate];
DisplayInfo=true;
DisplayError=true;
DisplayStepSize=true;
DisplayFinalResult=true;
DisplayOptions=[DisplayInfo ...
DisplayError ...
DisplayStepSize ...
DisplayFinalResult];
OptimizationMethod=1;
% 0: Backpropagation
% 1: Hybrid
fis=anfis([TrainInputs TrainTargets],fis,TrainOptions,DisplayOptions,[],OptimizationMethod);
%% Apply ANFIS to Data
Outputs=evalfis(Inputs,fis);
TrainOutputs=Outputs(TrainInd,:);
TestOutputs=Outputs(TestInd,:);
%% Error Calculation
TrainErrors=TrainTargets-TrainOutputs;
TrainMSE=mean(TrainErrors.^2);
TrainRMSE=sqrt(TrainMSE);
TrainErrorMean=mean(TrainErrors);
TrainErrorSTD=std(TrainErrors);
TestErrors=TestTargets-TestOutputs;
TestMSE=mean(TestErrors.^2);
TestRMSE=sqrt(TestMSE);
TestErrorMean=mean(TestErrors);
TestErrorSTD=std(TestErrors);
%% Plot Results
figure;
PlotResults(TrainTargets,TrainOutputs,'Train Data');
figure;
PlotResults(TestTargets,TestOutputs,'Test Data');
figure;
PlotResults(Targets,Outputs,'All Data');
if ~isempty(which('plotregression'))
figure;
plotregression(TrainTargets, TrainOutputs, 'Train Data', ...
TestTargets, TestOutputs, 'Test Data', ...
Targets, Outputs, 'All Data');
set(gcf,'Toolbar','figure');
end
figure;
gensurf(fis, [1 2], 1, [30 30]);
xlim([min(Inputs(:,1)) max(Inputs(:,1))]);
ylim([min(Inputs(:,2)) max(Inputs(:,2))]);
