Merge multi-row values to single row in table for classification training

Hi, i am stuck in preparing data for training. i have a table like

 P      Q       
 5  4  
 3  3  
 2  3  
 1  4  
 1  5  
 ...

i wish to create another table from it like:

 P(t-5) Q(t-5)  P(t-5)  Q(t-5)  P(t-5)  Q(t-5)  P(t-5)  Q(t-5)  P(t)    Q(t)     
 0  0  0  0  0  0  0  0  5  4     
 0  0  0  0  0  0  5  4  3       3
 0  0  0  0  5  4  3       3       2       3
 0  0  5  4  3       3       2       3       1       4
 5  4  3       3       2       3       1       4       1       5
 ...

For a big dataset, i wonder except for for-loop, if there is any other ways to compute it? i actually want to feed this small section(curve) data to train my classification. i wonder if this is a good idea.

Answer 1

[Eric Tao]

Assume your original table is called 'data', and results will be stored in a new table called 'new', then run:

new = table();
new.P_minus4 = [zeros(4,1);data.P(1:end-4)];
new.Q_minus4 = [zeros(4,1);data.Q(1:end-4)];
new.P_minus3 = [zeros(3,1);data.P(1:end-3)];
new.Q_minus3 = [zeros(3,1);data.Q(1:end-3)];
new.P_minus2 = [zeros(2,1);data.P(1:end-2)];
new.Q_minus2 = [zeros(2,1);data.Q(1:end-2)];
new.P_minus1 = [zeros(1,1);data.P(1:end-1)];
new.Q_minus1 = [zeros(1,1);data.Q(1:end-1)];
new.P = data.P;
new.Q = data.Q;

And you will get what you want. If you want the code to be more concise, run this lines:

new = table();
for i = 4:-1:1
  m = num2str(i);
  eval(['new.P_minus',m,' = [zeros(',m,',1);data.P(1:end-',m,')];']);
  eval(['new.Q_minus',m,' = [zeros(',m,',1);data.Q(1:end-',m,')];']);
end
new.P = data.P;
new.Q = data.Q;