Time Series with unique column values

I have a csv data file that I've attached. I am looking to split the data into 5 minute intervals and find the mean of each unique column value under 'label' (i.e. 'cpu', 'memory', 'disk') within that interval.

I understand that the discretize function of matlab can bin my datetime data with respect to 'time', however the problem I am having sorting the unique column vectors of 'label' for the bins in order to categorize them individually with respect to their 'label'.

Essentially, I am looking for an output that will resemble something similar to: 'cpu_mean: xxx memory_mean: xxx disk_mean: xxx'

for their respective intervals. But i am not sure how to address the unique column values of 'label'.

Answer 1

[jonas]

Something like this

data=readtable('sample09.csv')
[G nms]=findgroups(data(:,1));
t = datetime(data{:,2},'InputFormat','yyyy-MM-dd''T''HH:mm:ss.SSSXXX','TimeZone','UTC')
t1=t(G==1);
TT=timetable(t1,data{G==1,3},data{G==2,3},data{G==3,3});
TT=sortrows(TT);
newtime=t1(end):minutes(5):t1(1);
TT_mean=retime(TT,newtime,'mean')
TT_mean.Properties.VariableNames=nms{:,1}'

             t1              cpu     disk     memory
    ____________________    _____    _____    ______

    01-Aug-2018 18:26:00    0.304    0.716    0.294 
    01-Aug-2018 18:31:00    0.714    0.686    0.252 
    01-Aug-2018 18:36:00    0.408    0.668    0.554 
    01-Aug-2018 18:41:00     0.45    0.642    0.516 
    01-Aug-2018 18:46:00    0.498    0.626    0.366 
    01-Aug-2018 18:51:00    0.568    0.592    0.286 
    01-Aug-2018 18:56:00    0.628    0.566    0.388 
    01-Aug-2018 19:01:00    0.382    0.552    0.322 
    01-Aug-2018 19:06:00     0.41    0.524    0.568 
    ...

This works if all three measurements are taken at the same times, i.e. for each time you have three values (cpu/disk/memory). This seems to be the case when I checked the data.