How to cluster data in a histogram ?
19 visualizaciones (últimos 30 días)
Mostrar comentarios más antiguos
Sim
el 3 de Nov. de 2022
Comentada: Sim
el 3 de Nov. de 2022
How to cluster data in a histogram ?
Desired output
Input
A = duration({'00:01:01'
'00:00:53'
'00:00:55'
'00:00:54'
'00:00:54'
'00:00:53'
'02:45:08'
'00:01:33'
'00:00:57'
'00:00:58'
'00:00:51'
'00:00:45'
'00:01:03'
'00:00:56'
'00:00:45'
'00:26:52'
'00:01:12'
'00:00:41'
'00:00:56'
'00:01:16'
'00:01:47'
'00:09:22'
'00:00:40'
'00:00:38'
'00:00:48'
'00:00:38'
'00:00:42'
'00:00:42'
'00:01:06'
'00:01:00'
'00:00:43'
'00:00:47'
'00:00:43'
'00:00:50'
'00:00:52'
'00:01:20'
'00:01:35'
'00:00:54'
'00:01:05'
'00:02:07'
'00:00:43'
'00:00:39'
'00:29:36'
'00:00:39'
'00:00:39'
'00:01:01'
'00:01:09'
'00:01:12'
'00:01:11'
'00:01:12'
'00:01:06'
'00:01:06'
'00:01:00'
'00:01:15'
'00:01:08'
'00:00:39'
'00:00:59'
'00:00:54'
'00:01:25'
'00:01:01'
'00:01:03'
'00:01:03'
'00:00:56'
'00:01:19'
'00:01:05'
'00:01:00'
'00:01:09'
'00:01:12'
'00:00:52'
'00:00:40'
'00:01:09'
'00:01:00'
'00:01:04'
'00:00:57'
'00:02:07'
'00:02:44'
'00:00:51'
'00:01:22'
'00:01:10'
'00:01:07'
'00:01:07'
'00:00:48'
'00:00:59'
'00:01:02'
'00:00:48'
'00:00:49'
'00:00:56'
'00:01:03'
'00:00:53'
'00:01:23'
'00:00:40'
'00:01:25'
'00:01:15'
'00:01:13'
'00:02:14'
'00:01:08'
'00:00:53'
'00:01:00'})
binLimits = [min(A) max(A)];
binWidth = duration('00:00:5');
[counts,binEdges] = histcounts(A, 'BinLimits',binLimits,'BinWidth',binWidth);
histogram(A,'binEdges',binEdges,'FaceColor','k','EdgeColor','k');
0 comentarios
Respuesta aceptada
Cris LaPierre
el 3 de Nov. de 2022
You need to define the criteria for what makes a cluster and what makes an outlier. Once you know how you will define that, then you can use rmoutliers to apply your criteria and remove the outliers.
To me, it looks like you want to use quartiles. I find viewing this with a boxchart is easiest. Note that MATLAB does not accept the duration data type for outlier detection. Use the minutes function to covert your durations into numeric values, and vice versa.
A = duration({'00:01:01';'00:00:53';'00:00:55';'00:00:54';'00:00:54';'00:00:53';'02:45:08';'00:01:33';
'00:00:57';'00:00:58';'00:00:51';'00:00:45';'00:01:03';'00:00:56';'00:00:45';'00:26:52';'00:01:12';
'00:00:41';'00:00:56';'00:01:16';'00:01:47';'00:09:22';'00:00:40';'00:00:38';'00:00:48';'00:00:38';
'00:00:42';'00:00:42';'00:01:06';'00:01:00';'00:00:43';'00:00:47';'00:00:43';'00:00:50';'00:00:52';
'00:01:20';'00:01:35';'00:00:54';'00:01:05';'00:02:07';'00:00:43';'00:00:39';'00:29:36';'00:00:39';
'00:00:39';'00:01:01';'00:01:09';'00:01:12';'00:01:11';'00:01:12';'00:01:06';'00:01:06';'00:01:00';
'00:01:15';'00:01:08';'00:00:39';'00:00:59';'00:00:54';'00:01:25';'00:01:01';'00:01:03';'00:01:03';
'00:00:56';'00:01:19';'00:01:05';'00:01:00';'00:01:09';'00:01:12';'00:00:52';'00:00:40';'00:01:09';
'00:01:00';'00:01:04';'00:00:57';'00:02:07';'00:02:44';'00:00:51';'00:01:22';'00:01:10';'00:01:07';
'00:01:07';'00:00:48';'00:00:59';'00:01:02';'00:00:48';'00:00:49';'00:00:56';'00:01:03';'00:00:53';
'00:01:23';'00:00:40';'00:01:25';'00:01:15';'00:01:13';'00:02:14';'00:01:08';'00:00:53';'00:01:00'});
B = minutes(A)
boxchart(B)
Here, outlers are indicated with the 'o' marker. Outliers are values that are more than 1.5 · IQR away from the top or bottom of the box. You indicate there should be 4, but we see the default definition classifies more, so you will need to use a custom definition.
I find using the Clean Outlier Data task in a live script is a quick and interactive way to find the desired threshold. Since tasks don't work, here, I'll just use rmoutliers with a threshold of 20 applied.
% Remove outliers
B= rmoutliers(B,"quartiles","ThresholdFactor",20);
% View results
boxchart(B)
% convert back to duration
A = minutes(B);
A.Format = 'hh:mm:ss'
% Original code for creating a histogram
binLimits = [min(A) max(A)];
binWidth = duration('00:00:5');
[counts,binEdges] = histcounts(A, 'BinLimits',binLimits,'BinWidth',binWidth);
histogram(A,'binEdges',binEdges,'FaceColor','k','EdgeColor','k');
Más respuestas (0)
Ver también
Categorías
Más información sobre Cluster Analysis and Anomaly Detection en Help Center y File Exchange.
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!