% student's project 2006 on Provlepsis lesson
% PREDICTION USING AN ---ANFIS--- MODEL  :PROJECT 6

close all  %clean the workspace
clear all
clc

 mydata=xlsread ('C:\Users\User\Desktop\Ziogas.xlsx', 'e1:e252'); % retrieves the E column data of range 1 to 252 from the data file 'atermondata'

%load agrasf
%mydata


tic
figure(1) % a view of data
plot (mydata)
xlabel('time'); ylabel('values')
title('Actual values')


%  Estimation of an ANFIS model
% prepare training data
%input data
tr=mydata(1:200) % TRAINING DATA
% input (k-2)
train=tr
train(length(train))=[]% removes the last row
train(length(train))=[]; %removes the second last row
length(train)


%input k-1

train1=tr % first input
train1(length(train1))=[]% removes the last row
train1(1)=[] % removes the first row
length(train1)

% output k
train2=tr %second input
train2(1)=[]  % removes the first row
train2(1)=[]  %removes the second row
length(train2)



trn_data=[train train1 train2 ] %(k-2) (k-1) (k) training data



%%%%%% Ploting TRAINING data as a scater plot%%%%%%%
figure(10)
subplot(1,2,1)
plot (train, train2, 'o')
xlabel ('input (k-2) ')
ylabel('ouput (k)')
title('Training data')
axis equal; axis square

subplot(1,2,2)
plot (train1, train2, 'o')
xlabel ('input (k-1) ')
ylabel('ouput (k)')
title('Training data')
axis equal; axis square




%preparing the evaluation(test) data
ev=mydata(201:end)% TESTING DATA

% input (k-2)
eval=ev
eval(length(eval))=[]; eval(length(eval))=[];
length(eval)

% input (k-1)
eval1=ev
eval1(length(eval))=[];
eval1(1)=[];
length(eval)

% input (k)
eval2=ev
eval2(1)=[]
eval2(1)=[];
length(eval2)


evaldata=[eval eval1]   %input (k-2) and  (k-1)

y2=eval2 %(k) output data for testing





% generate FIS matrix
epoch_n=400
mf_n=[4 4];

%mf_type='gbellmf';   %type of membership function
%mf_type='trimf';   %the parameter b>c
%mf_type='gauss2mf'
mf_type='gaussmf' %
%mf_type='smf' % unsupported
%mf_type='trapmf' % the parameter b>c
%mf_type='zmf'  %unsupported
%mf_type='pimf' % run problem

in_fismat=genfis1(trn_data, mf_n, mf_type);


% start training
ss=0.1;

[m_anfis trn_error step_size ] = ...
    anfis(trn_data, in_fismat, [epoch_n nan ss nan nan], [1,1,1,1]);



figure('name', ['ANFIS: time series prediction'])

subplot(211);
tmp=[trn_error ];
plot(tmp);
title('Error Curves');
axis([0 epoch_n min(tmp(:)) max(tmp(:))]);
xlabel('epochs')
ylabel('RMSE')
legend('Training Error');


subplot(212);
plot(step_size);
xlabel('epochs')
title('Step Size');



% plot the initial membership functons
figure (20)
subplot(2,1,1)
[mfx, mfy]=plotmf(in_fismat, 'input', 1);
plot(mfx, mfy);
title('(a) Initial MFs on input')
axis([-inf inf -inf inf]);
subplot(2,1,2)
[mfx, mfy]=plotmf(in_fismat, 'input', 2);
plot(mfx, mfy);
title('(b) Initial MFs on input')
axis([-inf inf -inf inf]);


% plot final MF's on x,y,z,u

figure (30)
subplot(2,1,1)
[mfx, mfy]=plotmf(m_anfis, 'input', 1);
plot(mfx, mfy);
title('(a) Final MFs on input')
axis([-inf inf -inf inf]);
subplot(2,1,2)
[mfx, mfy]=plotmf(m_anfis, 'input', 2);
plot(mfx, mfy);
title('(b) Final MFs on input')
axis([-inf inf -inf inf]);




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%************


figure(40)
plotfis(in_fismat)

figure(50)
plotfis(m_anfis)

showrule(m_anfis)


anfisedit (m_anfis)
surfview(m_anfis)


%%%%%%%%%%%%%%%%%%%%%%%%%% IN SAMPLE EVALAUTION %%%%%%%%%%%%
insample_data=trn_data(:,1:2)
insaple_output=trn_data(:,3)
yhat_anfis_insample=evalfis(insample_data, m_anfis);

figure (60)
plot(insaple_output(end-30:end), 'b-s'), hold, plot(yhat_anfis_insample(end-30:end), 'r-x');
legend('actual values','ANFIS prediction values')
xlabel('time')
ylabel('values')
title('Actual values and ANFIS prediction in sample')


figure (70)
plot(insaple_output-yhat_anfis_insample)
xlabel('time')
ylabel('error')
title('Prediction in sample errors')



%Root Mean Square Error (RMSE)
RMSE_anfis_insample=sqrt(norm(insaple_output-yhat_anfis_insample)^2/length(insaple_output-yhat_anfis_insample))





%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%prediction by ANFIS model out of sample

%%%%%%%%%%%%%%%%%%%%%%%%%%% ANFIS EVALUATION %%%%%%%%%%%%%%%%%
input=evaldata


%%%%%% Ploting EVALUATION data as a scater plot%%%%%%%
figure(90)
subplot(1,2,1)
plot (eval, eval2, 'o')
xlabel ('input (k-2)')
ylabel('ouput (k)')
title('Evaluating data')
axis equal; axis square

subplot(1,2,2)
plot (eval1, eval2, 'o')
xlabel ('input (k-1)')
ylabel('ouput (k)')
title('Evaluating data')
axis equal; axis square



%%%%%%%%% Evaluation of Anfis out of sample
yhat_anfis=evalfis(input, m_anfis);


adapt_input=y2;
length(adapt_input);
length(yhat_anfis);
result=[adapt_input yhat_anfis (adapt_input-yhat_anfis)];
er_anfis=adapt_input-yhat_anfis; %error


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
figure (100)
plot(adapt_input(end-30:end), 'b-s'), hold, plot(yhat_anfis(end-30:end), 'r-x');
legend('actual values','ANFIS prediction values')
xlabel('time')
ylabel('values')
title('Actual values and ANFIS out of sample prediction')


figure (120)
plot(adapt_input-yhat_anfis)
xlabel('time')
ylabel('error')
title('Prediction errors')




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% error measures calculation


MSE_anfis=(1/length(yhat_anfis))*norm(er_anfis)^2

%Root Mean Square Error (RMSE)
RMSE_anfis=sqrt(norm(er_anfis)^2/length(er_anfis))

%Mean Absolute Error (MAE)
MAE_anfis=(1/length(er_anfis))*sum(abs(er_anfis))

%Mean Absolute percentage Error (MAPE)
MAPE_anfis=(100/length(er_anfis))*sum(abs(er_anfis)./abs(adapt_input))

toc  % the calculation time in seconds
runing_minute_time=toc/60
%end