FITNET

clc
clear all
close all
format long
out = csvread('Data_1000points_sampled_every5.csv');%importing the data file
u= [out(:,1) out(:,2)]'; % matrix containing u1 and u2
xo = out(:,3)'; %matrix containing x1
time=out(:,4)'; % time stamps

[row_u,col_u]=size(u); %getting size of dtatset
[row_xo,col_xo]=size(xo);



num_of_elements_in_validation=100;% Setting size of validation set

% creating training set
training=[];
training=[u(1,1:col_u-num_of_elements_in_validation);u(2,1:col_u-num_of_elements_in_validation);xo(1,1:col_u-num_of_elements_in_validation)];
[row_training,col_training]=size(training);
training_time=[time(1,1:col_u-num_of_elements_in_validation)];

% creating validation set
validation=[];
validation=[u(1,col_u-num_of_elements_in_validation+1:col_u);u(2,col_u-num_of_elements_in_validation+1:col_u);xo(1,col_u-num_of_elements_in_validation+1:col_xo)];
[row_validation,col_validation]=size(validation);
validation_time=[time(1,col_u-num_of_elements_in_validation+1:col_u)];


weighted_in =[];%matrix to store  u1 and u2 and x1  values from orignal dataset
weighted_out=[];%stores value of x1 of the point that is next, "not fed into net"

column_num1=1;% keeping track of column index
sim_ans_validation=[];% stores the simulated value of x1 generated by sim function



% how many elements we want to look back on
num_of_elements=4;
for n=1:20
    num_of_elements=n;

column_num1 = 1;
column_num=1;
row_num=1;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
weighted_in =[];%matrix to store  u1 and u2 and x1  values from orignal dataset
weighted_out=[];%stores value of x1 of the point that is next, "not fed into net"

column_num1=1;% keeping track of column index
sim_ans_validation=[];% stores the simulated value of x1 generated by sim function
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


for i=1:col_training-num_of_elements % outer loop
    row_num=1;
    for j=i:i+num_of_elements-1% inner loop
        %in this loop we generate our smaller datasets from training set and store them in a matrix
        
        weighted_in(row_num,column_num)=training(1,j);%u1 values
        row_num=row_num+1;
        weighted_in(row_num,column_num)=training(2,j);%u2 values
        row_num=row_num+1;
        weighted_in(row_num,column_num)=training(3,j);% x1 values coressponding to u1 and u2 values 
        %in the above statement
        
        row_num=row_num+1;
    end
    weighted_out(1,column_num)=xo(1,i+num_of_elements);
    column_num=column_num+1;
    
    

end


% feeding it into the RBF
%net=newrbe(weighted_in,weighted_out);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
net = fitnet(10);
     net = train(net,weighted_in,weighted_out);
     view(net)
     y = net(weighted_in);
perf = perform(net,weighted_out,y);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% now we must create a similar matrix for the validation set
weighted_in_valid=[];
row_num=1;
column_num=1;


for z=1:col_validation-num_of_elements% outer loop
    
    row_num=1;
    for y=z:z+num_of_elements-1% inner loop
        %in this loop we generate our smaller datasets from validation set
        
        weighted_in_valid(row_num,column_num)=validation(1,y);
        row_num=row_num+1;
        weighted_in_valid(row_num,column_num)=validation(2,y);
        row_num=row_num+1;
        weighted_in_valid(row_num,column_num)=validation(3,y);
        row_num=row_num+1;
    end
    weighted_out_valid(1,column_num)=xo(1,z+num_of_elements);
    column_num=column_num+1;
end


% sim_ans_training(1,:)= sim(net, weighted_in);%running a simulation for the training set
% sim_ans_training(2,:)= training_time(1,num_of_elements+1:end);%storing corressponding timestamps
% 
% 
% 
% figure
% plot( training_time,training(3,:),'g',sim_ans_training(2,:),sim_ans_training(1,:),'b')
% legend("Actual training data","simulation for training ")
% title('Actual data and simulation of training set ')
% 
% 
% %now we run a simulation using the validation set values

sim_ans_validation(1,:) = sim(net,weighted_in_valid);% storing the simulated value
sim_ans_validation(2,:)=validation_time(1,num_of_elements+1:end);% storing the corresponding timestep


error_validation=[];

for i=1:col_validation-num_of_elements
error_validation(1,i)= abs(sim_ans_validation(1,i)-validation(3,num_of_elements+i))/validation(3,num_of_elements+i);
end
mean_error_validation(1,n)= mean(error_validation,'all')*100;
mean_error_validation(2,n)= n;
end

figure 
plot(mean_error_validation(2,:), mean_error_validation(1,:))
figure
plot(validation_time,validation(3,:),'g',sim_ans_validation(2,:), sim_ans_validation(1,:));
legend('actual data of  validation set', ' simulated data for validation set');
title(' actual and simulated for validation set');