creating for and while loop for an array data

array_data=30 x 1 % double
sliding_window=10;
time_sliding_window = 1:1:10;   % constant for time_sliding_window 
array_polynomial = polyfit(time_sliding_window,array_data(1:10)',2);          %2nd degree polynomial coefficients
array_predicted=polyval(array_polynomial,time_10);    %predicted values from polynomial
for i=1:sliding_window  
    if abs(array_data(i)-array_predicted(i)) > 1.0 
        error_row_1(i)=i;
        if exist('error_row_1')==1
            break
        end
    end
end
% assume error_row_1=5
if exist('error_row_1')==1 
    % reset parameters
    array_polynomial = polyfit(time_sliding_window,array_data(6:15)',2); 
    array_predicted=polyval(array_polynomial,time_sliding_window);    %predicted values from polynomial
    % run again the algorithm
for i=1:error_row_1+1:error_row_1+sliding_window  % starts after the error_row to another 10 data
if abs(array_data(i)-array_predicted(i)) > 1.0
        error_row_2(i)=i;
        if exist('error_row_2')==1
            error_row=[error_row_1;error_row_2];
    break
     end
    end
end
end

if exist('error_row_1')==0
% reset parameters
array_polynomial = polyfit(time_sliding_window,array_data(11:20)',2);  
array_predicted=polyval(array_polynomial,time_10);    %predicted values from polynomial
for i=1:sliding_window  
    if abs(array_data(i)-array_predicted(i)) > 1.0 
        error_row_2(i)=i;
    if exist('error_row_2')==1
            break
        end
    end
end
end

.......

In these above codes,

 if abs(array_data(i)-array_predicted(i)) > 1.0

this code needs to run 10 by 10 within the size of array_data (384). If this if condition happens, error_row should be created, then this if condition needs to be updated started from error_row+1 to error_row+10. If this if condition doesn't happen, array_data needs to be increased by 10 (from array_data(1:10) to array_data(11:20) until it reaches array_data(21:30).

How can I write a compact code for the above computations?

[Jan]

This piece of code is meaningless:

error_row_1(i)=i;
if exist('error_row_1')==1
    break
end

Reduce this to:

error_row_1(i)=i;
break

Or even better: Replace

for i=1:sliding_window  
    if abs(array_data(i)-array_predicted(i)) > 1.0 
        error_row_1(i)=i;
        if exist('error_row_1')==1
            break
        end
    end
end

by

error_row_1 = find(abs(array_data - array_predicted) > 1.0)

The other exist() calls are not useful also.

What does this mean: "this code needs to run 10 by 10 within the size of array_data (384)"?

[sermet OGUTCU]

assuming array_data=30 x 1, and sliding_window=10; if, if abs(array_data(i)-array_predicted(i)) > 1.0 condition doesn't happen at all, 10 batches of data for for loop should be; array_data(1:10), array_data(11:20), array_data(21:30). For example;

array_polynomial = polyfit(time_sliding_window,array_data(1:10)',2);          %2nd degree polynomial coefficients
array_predicted=polyval(array_polynomial,time_10);    %predicted values from polynomial
for i=1:sliding_window  
    if abs(array_data(i)-array_predicted(i)) > 1.0 
        error_row_1(i)=i;
        if exist('error_row_1')==1
            break
        end
    end
end
array_polynomial = polyfit(time_sliding_window,array_data(11:20)',2);  
array_predicted=polyval(array_polynomial,time_10);    %predicted values from polynomial
for i=1:sliding_window  
    if abs(array_data(i)-array_predicted(i)) > 1.0 
        error_row_2(i)=i;
        if exist('error_row_2')==1
            break
        end
    end
end
end
array_polynomial = polyfit(time_sliding_window,array_data(21:30)',2);  
array_predicted=polyval(array_polynomial,time_10);    %predicted values from polynomial
for i=1:sliding_window  
    if abs(array_data(i)-array_predicted(i)) > 1.0 
        error_row_2(i)=i;
        if exist('error_row_2')==1
            break
        end
    end
end
end

Answer 1

[Voss]

I believe this has the behavior you are after.

I'm not sure what should happen when you get to the end of array_data, i.e., maybe you want to use a partial window instead of always requiring 10 samples. As written, this code uses a fixed window size of 10 samples.

%%%%%% create random data, 30-by-1
mean_data = 10;
range_data = 3;
array_data = mean_data-range_data/2+range_data*rand(30,1);
%%%%%% end create data


N_data = numel(array_data);
window_size = 10;


%%%%%% plotting setup to show behavior
figure();
ylim(mean_data+range_data*[-1 1]);
hold on
colors = 'brgkmcy';
markers = 'o.xs';
N_colors = numel(colors);
N_markers = numel(markers);
count = 0;
%%%%%% end plotting setup


t = 1:window_size;

while t(end) <= N_data
    
    
    %%%%%% display stuff to command window
    disp('current t:');
    disp(t);
    %%%%%% end display stuff
    
    
    array_polynomial = polyfit(t,array_data(t).',2); % 2nd degree polynomial coefficients
    array_predicted = polyval(array_polynomial,t); % predicted values from polynomial
    
    
    %%%%%% plotting stuff
    count = count+1;
    color = colors(mod(count-1,N_colors)+1);
    marker = markers(mod(count-1,N_markers)+1);
    plot(t,array_data(t),[marker color],t,array_predicted,color);
    temp_data = [array_data(t); array_predicted(:)];
    temp_lim = [min(temp_data) max(temp_data)];
    line(t([1 end end 1 1]),temp_lim([1 1 2 2 1]), ...
        'Color',color,'LineWidth',2,'LineStyle','--');
    %%%%%% end plotting stuff
    
    
    error_row = t(1) - 1 + find(abs(array_data(t).'-array_predicted) > 1, 1);
    if isempty(error_row)
        t = t+window_size;
    else
        t = error_row+(1:window_size);
    end
    
    
    %%%%%% display stuff to command window
    disp('error_row:');
    disp(error_row);
    %%%%%% end display stuff
    
    
end

current t:

     1     2     3     4     5     6     7     8     9    10

error_row:

     8

current t:

     9    10    11    12    13    14    15    16    17    18

error_row:

current t:

    19    20    21    22    23    24    25    26    27    28

error_row:

    20

current t:

    21    22    23    24    25    26    27    28    29    30

error_row:

    25

[sermet OGUTCU]

Dear @Editor, thank you so much for this effective solution.

[Voss]

You're welcome!