outlier detection

what would be the simplest way for outlier detection using knime?

Well, it depends what kind of outlier detection you want to use... The simplest option is normalizing the columns you need with Z-score and remove those rows that are outside of certain distance from 0 with a row filter. You can denormalize later, or use a Reference Row Filter on the original table.

If you do not mind a manual step, the Conditional Box Plot/Box Plot nodes allow you to visualize the (robust) outliers, select and HiLite them and after the filter them with the HiLite Filter node.

For more complex outlier methods, you might need other nodes.

1 Like