Hi all,

Is there a node or measure in KNIME through which I can calculate the percentage of a point (or a document) being an outlier?

In other words, I have some points (or documents) and i want to prove that these points or these documents are outliers. So, I need a node or measurement for outliers such as for example Grabb's test for outliers or any other measurement.

Thanks in advance

Hi singing bird,

Grubbs' test is used to detect a single outlier in a univariate data set that follows an approximately normal distribution (retrieved from http://www.itl.nist.gov/div898/handbook/eda/section3/eda35h1.htm). Actually, this is not available in KNIME Analytics Platform. One option to detect outliers in KNIME Analytics Platform would be using a Box Plot which you can find in the Views category, or using the JavaScript BoxPlot available in KNIME Labs/JavaScript Views. After using the box plot, you can use a Row Filter node to filter all out the outliers. Another possible way would be to use the integration with R, and so use the R snippet node. There are some packages really useful for outlier detection like the mvoutlier package or the outliers package.

Hope that helps,

Best,

Vincenzo