What's New in KNIME 2.9

Highlights  

 - Rule Based Framework Extended and Improved
 - Column Selection Framework Extended and Improved
 - Better Model Training and Evaluation
 - Workbench Monitoring, Performance, and Administration
 - Data Generation
 - Other New Features and Nodes
 - Complete changelog File

 

Rule Based Framework Extended and Improved

New Rule Engine Node

 

New more powerful, more flexible, and more user-friendly Rule Engine Node.

The configuration window of the new Rule Engine node follows now the same GUI as the String Manipulation node and as the Math Formula node. This new User Interface allows for an easier inclusion of flow variables and data columns in the rule editing frame and for replacing existing columns in the output data table.

PMML Ruleset Editor

New Node!

This new node is based on the framework of the new Rule Engine node (see above).

It works like the  new Rule Engine node, producing the resulting values as well as the rule set as a PMML model at the output ports.

Rule Engine Variable

New Node!

This new node is based on the framework of the new Rule Engine node (see above).

It works similarly to the new Rule Engine node but on flow variables instead of data columns.

Rule Based Row Filter/Splitter

Two New Nodes!

These two new nodes follow the new Rule Engine node framework.

They filter/split data rows on the basis of the rule set defined in the configuration window.

Note that for this node the rule output must be TRUE or FALSE.

 

 

 

Column Selection Framework Extended and Improved

New Selection Criteria in Column Filter Node and similar nodes

 

The Column Filter configuration window now allows to select the input data columns:
  - manually
  - by name based on a RegEx / Wildcard expression
  - by type

This extended framework can be found in the Column Filter node as well as in all other nodes using the same column selection framework.

 

 

Better Model Training and Evaluation

Unified Format for Predictor Nodes

 

All predictor nodes now:

- show a similar configuration window
- produce output data tables with the same format
- can compute class probabilities on demand

Extended Linear Regression Output

 

The Linear Regression Learner node now has an additional data output port, which outputs the regression coefficients and p-values.

Numeric Scorer

New Node!

This new node scores predictions for numerical targets, by computing a number of error  measures, such as R², absolute and squared errors and the mean signed difference.

Weka 3.7

 

Integration with Weka has been upgraded to the latest version of Weka, Weka 3.7.

 

 

 

Workflow Monitoring, Performance, and Administration

Send Email

New Node!

The Send Email node sends an email to a given email address. It can be used to send updates about the workflow execution status.

Note. The node sends an email only if executed. If the workflow encounters an error before the Send Email node, you will receive no email.

Save Workflow

New Node!

This new node saves the workflow upon execution.

Save As ... New File Menu Option

 

New option "Save As ..." in the File menu to save the selected workflow under a different name.

TextProcessing Speed Up

 

Faster implementation of the text processing package.

Some nodes, like the tagger nodes and the string to Document node, can now be parallelized via the Concurrency tab or an additional option in the Options tab.

Easier Extension Installation

 

All existing KNIME update sites are now available in the Preferences menu item (File -> Preferences -> Install/Update -> Available Software Sites).

Just enable them, to perform a KNIME installation update or to install additional extensions from the File menu.

Context Extractor Node

New Node!

The node Extract Context Properties makes some workflow context related properties available at the variable output port, including current user name, workflow name, and working directory.

 

 

Data Generation

Counter Generation

New Node!

The Counter Generation node adds a counter column to the input data table.

The counter starting point (Min. Value) and the step size (Scale Unit) can be defined in the configuration window.

Time Series Generator

New Node!

The Time Series Generator node generates a time series as an additional column in the input table, by repeating a pattern wave: saw tooth wave, sine wave, square wave, or triangle wave.

 

 

Other New Features and Nodes

   

Cross Joiner

New Node!

Performs a cross join of two tables.

Each row of the first table is joined with each row of the second table.

Pair Extractor

New Node!

"Unpivot" a distance matrix by extracting all pairwise distances from a distance matrix column.  

Distance Matrix to Network

New Node!

Insert features from a Distance Matrix column into a network.  Thresholded distances can be used as weights for edges.  

Flow Variables and Missing Values

 

Nodes TableRow To variable and TableRow To Variable Loop Start now include a strategy to deal with missing values. 

In the configuration window you can provide substitute values for possible missing values.

 

Many more small improvements have been made under the hood - please refer to the changelog