The following node is available in the Open Source KNIME predictive analytics and data mining platform version 2.7.1. Discover over 1000 other nodes, as well as enterprise functionality at http://knime.com.

Backward Feature Elimination Filter

This node takes a model built with a feature elimination loop and lets you choose the column you want to include in the output table. The dialog will show you all computed levels of the feature elimination together with the error rates. You may then either select one level ("manual selection") or you specify an error threshold and then the level with the fewest features that has a prediction error below the threshold is automatically selected. In any case all columns from the input table that are not present in the selected level are filtered from the input table.
You may optionally include the target column. Note that the column names must be the same as the ones used for the elimination loop. If they are not, rename them first.

Dialog Options

Include target column
If checked, that target column is included in the output table, otherwise it is also filtered out (if it exists)
Select features manually
By selecting this option you can choose a set of features in the level table below.
Select features automatically by error threshold
By selecting this option you can set a prediction error threshold.
Prediction error threshold
Enter a prediction error threshold here. The level with the fewest number of features that is below the threshold will be selected automatically.
Level table
Shows the levels (i.e. number of features) and the corresponding error rates. You may click on a row and the column included in this level are selected in the...
Included columns list
Show all columns that will be included in the output table.

Ports

Input Ports
0 A backward feature elimination model
1 Any datatable that should contain the same columns as used in the elimination loop
Output Ports
0 The input table with some columns filtered out
This node is contained in KNIME Base Nodes provided by KNIME GmbH, Konstanz, Germany.