How do i insert seconds in a timeseries data when the frequency of the data is inconsistent?

I have a timeseries data as:

And i need to insert seconds on this data in way that it depends upon the number of datapoints the minutues has. For example, 9/5/2021 11:30:00 has 5 datapoints so every data will be increased by {60/5= 12 seconds), 11:31:00 has 3 data points so, every data is increased by {60/3=20 seconds).

 Input_data= 
  9/5/2021 11:30:00  
  9/5/2021 11:30:00  
  9/5/2021 11:30:00  
  9/5/2021 11:30:00  
  9/5/2021 11:30:00  
  9/5/2021 11:31:00
  9/5/2021 11:31:00
  9/5/2021 11:31:00
  9/5/2021 11:32:00
  9/5/2021 11:32:00
  9/5/2021 11:32:00
  9/5/2021 11:32:00
  
Final_output= 
  9/5/2021 11:30:00  
  9/5/2021 11:30:12  
  9/5/2021 11:30:24  
  9/5/2021 11:30:36  
  9/5/2021 11:30:48  
  9/5/2021 11:31:00
  9/5/2021 11:31:20
  9/5/2021 11:31:40
  9/5/2021 11:32:00
  9/5/2021 11:32:15
  9/5/2021 11:32:30
  9/5/2021 11:32:45 

Answer 1

[Star Strider]

Try this —

Input_data = ['9/5/2021 11:30:00' 
              '9/5/2021 11:30:00'  
              '9/5/2021 11:30:00'  
              '9/5/2021 11:30:00'  
              '9/5/2021 11:30:00'  
              '9/5/2021 11:31:00'
              '9/5/2021 11:31:00'
              '9/5/2021 11:31:00'
              '9/5/2021 11:32:00'
              '9/5/2021 11:32:00'
              '9/5/2021 11:32:00'
              '9/5/2021 11:32:00'];
 
TV = datetime(Input_data, 'InputFormat','MM/dd/yyyy HH:mm:ss');         % Convert To 'datetime' Array
[TVu,~,Idx] = unique(TV,'stable');                                      % Unique Values In Original Order
Counts = accumarray(Idx,1);                                             % Count Unique Occurrences

for k = 1:numel(Counts)
    ss{k,:} = (0:Counts(k)-1).'*60/Counts(k);                           % Create 'seconds' Column Vector
end
cs = seconds(cell2mat(ss));                                             % Convert Numeric Values To 'duration' 'seconds'

TV = TV + cs                                                            % Add 'seconds' To Original Vector To Produce Desired Result

TV = 12×1 datetime array
   05-Sep-2021 11:30:00
   05-Sep-2021 11:30:12
   05-Sep-2021 11:30:24
   05-Sep-2021 11:30:36
   05-Sep-2021 11:30:48
   05-Sep-2021 11:31:00
   05-Sep-2021 11:31:20
   05-Sep-2021 11:31:40
   05-Sep-2021 11:32:00
   05-Sep-2021 11:32:15
   05-Sep-2021 11:32:30
   05-Sep-2021 11:32:45

I am not certain that this is robust enough to work with any ‘Input_data’ vector, however it works with the example provided.

(To understand how it works, remove the ending semicolons to see the interim results.)

.

Answer 2

[dpb]

indx=[0;find(minutes(diff(TT.Time)));height(TT)];       % find change locations indices
N=diff(indx);                                           % number observations/group
secs=60./N;                                             % number seconds differential in group
dsecs=arrayfun(@(s,n)s*[(0:n-1).'],secs,N,'UniformOutput',false);   % compute the vector of addends
dsecs=vertcat(dsecs{:});
TT.Time=TT.Time+dsecs;                                  % fixup the time field

Applied to your sample, this results in--

>> TT =
  12×1 timetable
            Time             x  
    ____________________    ____
    05-Sep-2021 11:30:00    0.18
    05-Sep-2021 11:30:12    0.34
    05-Sep-2021 11:30:24    0.21
    05-Sep-2021 11:30:36    0.51
    05-Sep-2021 11:30:48    0.91
    05-Sep-2021 11:31:00    0.63
    05-Sep-2021 11:31:20    0.10
    05-Sep-2021 11:31:40    0.39
    05-Sep-2021 11:32:00    0.05
    05-Sep-2021 11:32:15    0.50
    05-Sep-2021 11:32:30    0.43
    05-Sep-2021 11:32:45    1.00
>> 

where I just created a dummy variable to have a non-empty TT.