Self-Tuning Association Rules for KNIME

For the time being, the STARK project provides the first public version of the Yacaree Associator node, which computes association rules under very restrictive conditions that attempt at reducing greatly the intuitive redundancy of the outcome.

You should be able to download the node into your KNIME through the usual means: Help, Install New Software, Available Software Sites, KNIME Community Contributions, Stark. The node will appear under the Mining and Item Sets / Association Rules headings of the node repository. We recommend to find as well other nodes about itemsets and association rules available at the KNIME Update Site (KNIME and Extensions), for comparisons purposes; but they are not necessary. Alternatively, a jar file named es.unican.knime.stark_[version/build id] is available through (the Sourceforge page of the elder brother of this project, the Python yacaree associator); it contains a slightly older build, with only cosmetic differences. In this case, simply copy it to the dropins folder of your KNIME installation.

The node expects as input a table including a collection column where each cell provides a transaction. Items can be integers but they can be strings as well. One way to obtain the column from the usual space-separated transactional data files is to use a File Reader node (which may need to be configured with "allow short lines" in the Advanced button and space as separator) and to connect it to a Create Collection Column node.

STARK: Self-Tuning Association Rules for KNIME started as a project of the Harvest programme of Pascal-2: a Network of Excellence of the FP7 Framework Programme of the European Union which supports a large number of cooperative research initiatives in Machine Learning and Statistical Modeling. Within it, the Harvest Programme aims at fostering cooperation of researchers to produce software tools. STARK has the goal of turning a number of research contributions related to Association Rule Mining into a self-tuning associator module of KNIME that brings the results of this research at the fingertips of any practitioner interested in using it. a link to the STARK original proposal appears below.

Participating personnel include:

Javier de la Dehesa (Universidad de Cantabria),
Tobias Kötter (KNIME and Universität Konstanz),
José L Balcázar (Universitat Politècnica de Catalunya, and earlier also Universidad de Cantabria),
Michael Berthold (KNIME and Universität Konstanz),
Diego García-Sáiz (Universidad de Cantabria), and
Cristina Tîrnauca (Universidad de Cantabria).

Source Code

The source code can be accessed at


The STARK nodes are relased under GPLv3 with additional permissions.


What are you looking for?