%function [datas,datad,datad1,M]=dscale
%DSCALE
%------
%   This subfunction scales data to value between 0 and 1
%
%   datas   = scaled data
%   data    = actual data before scaling
%   min     = actual data at their minimum
%   max     = actual data at their maximum

load dat9features.txt;
newdata=dat9features;
%bearing=bearing(randperm(212),1:3); % Randomize
[R Q] = size (newdata);
P1 = newdata(:,1);
P2 = newdata(:,2);
P3 = newdata(:,3);
P4 = newdata(:,4);
P5 = newdata(:,5);
P6 = newdata(:,6);
P7 = newdata(:,7);
P8 = newdata(:,8);
P9 = newdata(:,9);
T = newdata(:,10);  %output data set


Pa = P1';
Pb = P2';
Pc = P3';
Pd = P4';
Pe = P5';
Pf = P6';
Pg = P7';
Ph = P8';
Pi = P9';
Ta = T';

for i=1:R
    data(1,i)=Pa(i);
    data(2,i)=Pb(i);
    data(3,i)=Pc(i);
    data(4,i)=Pd(i);
    data(5,i)=Pe(i);
    data(6,i)=Pf(i);
    data(7,i)=Pg(i);
    data(8,i)=Ph(i);
    data(9,i)=Pi(i);
    data(10,i)=Ta(i);
    end

[N M]=size(data);
for i=1:N
    max(i)=data(i,N);
    min(i)=data(i,N);
    for j=1:M
        if data(i,j)>max(i)
           max(i)=data(i,j);
        end
        if data(i,j)<min(i)
           min(i)=data(i,j);
        end
    end
    datas(i,:)=(data(i,:)-min(i))/(max(i)-min(i)); % no delay
    %datad(i,1:M-1)=datas(i,2:M);    % 1 delayed term
    %datad1(i,1:M-2)=datas(i,3:M);    % 2 delayed term
end


X=M;
% Training data
input(1,1:X)=datas(1,1:X);    % column P1
input(2,1:X)=datas(2,1:X);    % column P1
input(3,1:X)=datas(3,1:X);    % column P1
input(4,1:X)=datas(4,1:X); 
input(5,1:X)=datas(5,1:X); 
input(6,1:X)=datas(6,1:X); 
input(7,1:X)=datas(7,1:X); 
input(8,1:X)=datas(8,1:X); 
input(9,1:X)=datas(9,1:X); 
output(1,1:X)=datas(10,1:X);   % target

[pn,meanp,stdp,tn,meant,stdt] = prestd(input,output);

%[A B] = size(TrainIn);
in= input;
out= output;
input1=in*1.05;
output1 =out*1.05;
input2=in*1.1;
output2 = out*1.1;
ptr=in; ttr=out;	    % Training
v.P=input1; v.T=output1;    % Validatiorn 
t.P=input2; t.T=output2;    % Testing


S1=5; % Number of nodes 
fnet=newff(minmax(input),[S1 1],{'tansig' 'purelin'},'trainlm');
fnet.trainparam.epochs=5000;				% Max epoch number 
fnet.trainParam.maxit= 5000;
fnet.trainParam.goal=1e-12;
fnet.trainParam.max_fail=1000;
fnet.trainParam.show=10;
fnet.iw{1,1}
fnet.b{1}
fnet=init(fnet);

[fnet,tr]=train(fnet,ptr,ttr,[],[],v,t)
plot(tr.epoch,tr.perf,tr.epoch,tr.vperf,tr.epoch,tr.tperf);
legend('Training','Validation','Test',-1);
ylabel('Squared error');
xlabel('Epoch');


an=sim(fnet,input); %simulation the network
e=output'-an';
perf=sse(e);
perf1=mse(e)

van = sim(fnet,input1);
ve=output1'-van';
verr1 = sse(ve);
verr2 = mse(ve)

tan = sim(fnet,input2);
te = output2' - tan';
terr = sse(te);
terr2 = mse(te)


a = poststd(an,meant,stdt);
for z=1
figure(i)
[m(z),b(z),r(z)] = postreg(a(z,:),output(z,:));
end

figure(1);
s=(1:48); %plot the graf data vs month(1986-2001)
plot(s,output(s),'r+-',s,an,'b.-')
title('Training ')
xlabel('Observation Data')
ylabel('Classification Output (Normalized)')
legend(' = actual',' = prediction')
  
figure(2);
s=(1:48); %plot the graf data vs month(1986-2001)
plot(s,output1(s),'r+-',s,van,'b.-')
title('Validation ')
xlabel('Observation Data')
ylabel('Classification Output (Normalized)')
legend(' = actual',' = prediction')

figure(3);
s=(1:48); %plot the graf data vs month(1986-2001)
plot(s,output2(s),'r+-',s,tan,'b.-')
title('Testing ')
xlabel('Observation Data')
ylabel('Classification Output (Normalized)')
legend(' = actual',' = prediction')