Feature Selection Filter

This node takes a model built with a feature selection loop as input and lets you choose the subset of columns you want to include in the output table. The dialog shows all computed subsets together with their scores. You can select a subset manually or specify a score threshold. If you specify a threshold, the smallest features set whose score meets the threshold requirement is selected. If the "Minimize Score" option was checked in the Feature Selection Loop End node, this is the smallest feature set with a score below the specified threshold; if the "Minimize Score" option was left unchecked in the Feature Selection Loop End node, this is the smallest feature set with a score above the specified threshold.

Options

Include static columns

If this option is checked, any static columns are included in the output table, otherwise they are filtered out.

Select features manually

By selecting this option you can choose a set of features in the table below.

Select best score

By selecting this option the best score is chosen. In case the score is being maximized, the best score is the highest score value, while if the score is being minimized the best score is the lowest score value.

Select features automatically by score threshold

By selecting this option you can set a prediction score threshold.

Prediction score threshold

Enter a prediction score threshold here. The smallest feature set with a score below the specified threshold (if the score is to be minimized) or above the specified threshold (if the score is to be maximized) is selected.

Feature set table

The feature sets are listed on the right side, their scores and the number of features in these sets on the left side. Clicking on a score on the left side highlights the corresponding feature set on the right side.

Views

This node has no views

Workflows

• 01_AnalyticsKNIME Hub
• 01_AnalyticsKNIME Hub
• 01_Bike_Restocking_Alert_Training_w_Feature_SelectionKNIME Hub
• 01_Feature_Elimination_Done_ManuallyKNIME Hub
• 01_Feature_Elimination_with_Naive_BayesKNIME Hub
• Show all 140 workflows

Links

• No links available

Developers