数据分析Matlab / Excel

Question

I have a data clustering problem. I have a series of events labeled by timestamps and I am trying to count how many events is in each 15 (also 30) minute blocks. I am doing this in excel with pivot table . I can manage the 15 minute block but the problem is I need that if there is an empty block I need zero in that block . Instead excel doesn't show that block at all.

So how do I make that empty blocks can appear?

Related question I am using this blocks to create a vector in matlab and so far I haven't figured easy way to do that. I am little bit struggling how to import easily results of pivot table from excel to matlab.

Sample input:

30/11/12 12:42 AM
30/11/12 12:47 AM
30/11/12 12:56 AM
30/11/12 1:01 AM
30/11/12 1:52 AM
30/11/12 1:57 AM
30/11/12 2:38 AM
30/11/12 2:39 AM
30/11/12 6:00 AM
30/11/12 6:09 AM
30/11/12 6:16 AM
30/11/12 6:23 AM
30/11/12 6:31 AM

The pivot table will give

12:30 1
12:45 2
1:00 1
1:45 2
2:30 2
6:00 2
6:15 2
6:30 1

The problem is that from this I want to create a vector(each 15 minutes for the whole day meaning 24*4 = 96 elements) that will have "1" if there was an event and "0" if there was no event.

So output would look like from 00:00 to 6:30.

Output:

 vector = (0,0,1,1;1,0,0,1;0,0,1,0;0,0,0,0;0,0,0,0;0,0,0,0;1,1,1) 

where semicolon devides each hour just to read it easier

How to tackle this? Any hints? Is this easier to tackle in Matlab but the timestamps there are not that easy as in excel.

Answer 1

I'm not sure how to fix your Excel problem. But here is how to do this in Matlab:

%Data
dateStrings = {...
    '30/11/12 12:42 AM' ...
    '30/11/12 12:47 AM' ...
    '30/11/12 12:56 AM' ...
    '30/11/12 1:01 AM' ...
    '30/11/12 1:52 AM' ...
    '30/11/12 1:57 AM' ...
    '30/11/12 2:38 AM' ...
    '30/11/12 2:39 AM' ...
    '30/11/12 6:00 AM' ...
    '30/11/12 6:09 AM' ...
    '30/11/12 6:16 AM' ...
    '30/11/12 6:23 AM' ...
    '30/11/12 6:31 AM' ...
    };
%Convert data into datenums.  This is Matlab's standard numeric date encoding.
%    in units of days, starting at year 0000.
dateNumbers = datenum(dateStrings, 'dd/mm/yy HH:MM PM');

%Parametrically define the boundaries where you want to count
aggregationInterval = 1/24/4;  %15 minutes, in days\
aggregationStart = datenum('2012-11-30 00:00','yyyy-mm-dd HH:MM');
aggregationStop = datenum('2012-11-30 03:00','yyyy-mm-dd HH:MM');

%Use parameters to construct a vector of counting boundaries
aggregationBoundaries = aggregationStart:aggregationInterval:aggregationStop;

%The function histc does all the work, and returns a vector of counts
counts = histc(dateNumbers, aggregationBoundaries);

%This creates a cell to give you something to look at,  Instead of "disp" you coult use "xlswrite" to put this back into Excel.
disp([...
    cellstr(datestr(aggregationBoundaries','yyyy-mm-dd HH:MM')) ...
    num2cell(counts)])

This displays

'2012-11-30 00:00'    [0]
'2012-11-30 00:15'    [0]
'2012-11-30 00:30'    [1]
'2012-11-30 00:45'    [2]
'2012-11-30 01:00'    [1]
'2012-11-30 01:15'    [0]
'2012-11-30 01:30'    [0]
'2012-11-30 01:45'    [2]
'2012-11-30 02:00'    [0]
'2012-11-30 02:15'    [0]
'2012-11-30 02:30'    [2]
'2012-11-30 02:45'    [0]
'2012-11-30 03:00'    [0]

If your dates are already in Excel, you can also look at xlsread to read the values into Matlab without any text formatting.