Outlier detection

Hello everyone,

How is it possible to detect the outliers from a table in Knime? My table have values like

"Category"       "title"         "Price".

i want the outliers from the Price column, but i doent find a good way :(.

 

thanks a lot for your help!

Hello Linko:

It´s very easy to detect outliers using a Box Plot which you can find in Data Views. In general an outlier is a data that is far from some distance from Q3 or Q1. In Knime a Box Plot show an outlier with a 0 and an extreme with a +.

Here I leave a link where you can find more information about box plot.

http://www.purplemath.com/modules/boxwhisk3.htm

After used a box plot, you can use a row filter to filter all outliers and extremes

Gabriel

CHILE

 

Hi there is also a node called outlier removal and comes with the High Content Screening Tools (trusted extension) and can be found in https://tech.knime.org/community/hcs-tools

Basically I went to File/Preferences/InstallUpdate/Available software Sites and added this URL http://update.knime.org/community-contributions/trusted/2.11. Then I went to Install Knime extensions and typed HCS. 

Once installed you will find the node under Community Nodes/HCS Tools/Pre Processing/Outlier Removal.

You can also check this post https://tech.knime.org/node/46536/view for another method

Hope it helps

Cheers

 

 

Hi,

another way could be to normalize the data to z-scores with the "normalizer" node . then, rows whose values (in absolute value) are over a threshold (that you will have to determine) can be considered outliers.

I hope this is useful

 

Giovanni