KNIME: Example Workflows

Below you will find some example workflows that describe some key features in KNIME. You may download the workflows ad import them into your workspace by right-clicking into the Workflow Projects view and select Import KNIME Workflow...

Predictive Models - Training a decision tree

This workflow shows the concept of predictive models inside KNIME. They are clearly visible by the blue connections between the learner and the accompanying predictor nodes. The scorer node shows a confusion matrix and some quality measures in its view. The Color Manager node at the beginning of the workflow colors the rows according to their class values (e.g.). The colors are then visible in all successing views, like e.g. the Scatter Plotter. Download flow (32kB)

Meta workflows - Cross Validation

Checking the stability of a trained model is usually verified by cross validation. KNIME offers a special node, a so-called meta node, for this. The node encapsulates an inner workflow that is executed several times, the results of each iteration are collected and presented at the outer meta node in the end. The first output port gives the error rates for each iteration, the second port holds the predicted class values for each row. Download flow (888kB)

External libraries - CDK

Integrating functionality of external (Java) libraries into KNIME is fairly easy. A good example is the seemless integration of CDK (Chemistry Development Kit). Quite a lot of its methods and algorithms are available as KNIME nodes, the 2D coordinate generator being most popular node as it allows to view the molecules as 2D structure diagrams. In order to use the example workflow you need to install the CDK plugin via the Update Manager. Download flow (11MB)

Hiliting - Discriminative fragments and their molecules

Hiliting is one of the nicest features KNIME offers. Hiliting some data points in one branch of the workflow hilites corresponding items in all other parts. This feature is not restricted to a 1:1 correspondence which can be demonstrated nicely with the MoSS node for frequent substructure mining. The node finds frequent discriminative fragments in a set of molecules. Hiliting the found fragments in the Interactive Table following the MoSS (Molecular Substructure Search) node will also hilite all the molecules in which the fragment occurs in the upper branch. You will need the Chemistry Addons plugin for loading this workflow.

For space reasons this workflow contains some nodes that are not executed yet. Thus if you want to see the hiliting feature you first have to execute the remaining nodes. Download flow (2.2MB)