Universität Konstanz | Informatik und Informationswissenschaft | ag-sapozhnikova | Software

Interestingness Measures

This node calculates different interestingness measures between attributes of the different datasets. The following interestingness measures are supported. 

Jaccard 
Confidence 
Phi-Coefficient 
All-Confidence 
Kulc 
Lift 
Cosine

Download this node.

(Place the downloaded file into the 'plugins' folder in the KNIME installation folder)

Hierarchical Metrics

This node calculates Hierarchical multi-label performance Metrics by using different metrics on the predictions of multi-label algorithms. It uses the following metrics as in "An Empirical Comparison of Flat and Hierarchical Performance Measures for Multi-Label Classification with Hierarchy Extraction":


Kiritchenko. 
Verspoor et al. 
Ipeirotis et al. 
Sun and Lim 
Cai and Hofmann 
Cesa-Bianchi et al. 
Wu et al. 
Nowak and Lukashevich 
Wang et al. 
Struyf et al. 
Resnik

Download this node.

(Place the downloaded file into the 'plugins' folder in the KNIME installation folder)

Hierarchical Measures

This node calculates different hierarchical interestingness measures such as Jaccard Expectation, Jaccard Difference and Cosine as described in doi:10.1109/TKDE.2014.2320722.


Download this node.

(Place the downloaded file into the 'plugins' folder in the KNIME installation folder)

Apriori Hierarchy

This node calculates Apriori hierarchies of the provided data set as in doi:10.1016/j.patcog.2010.09.010. It outputs the rules of the provided data set, and these rules could be further used for example for drawing Hierarchies between different attributes of a data set. 

Download this node.

(Place the downloaded file into the 'plugins' folder in the KNIME installation folder)

Metric Visualization

This node visualizes the data in the form of a tree. The hierarchy and information visualized by the node is based on the selected interestingness measure, such as Jaccard Expectation, Jaccard Difference etc. The hierarchy could be seen in the form of a tree and its description in the form of a table. The further information such as support, support (AUB) etc could be seen when a particular node is selected from the tree.

Download this node.

(Place the downloaded file into the 'plugins' folder in the KNIME installation folder)

Meka Classifiers Package

The KNIME Meka Classifiers Package contains the node generator, the predictor and some other utils to classify multi-label datasets with KNIME.

The node generator give the option to select a classifier algorithm and to learn a model from input data.

The Meka Predictor takes a model generated in a meka node and classifies the test data at the inport.

Download this node.

(Place the downloaded file into the 'plugins' folder in the KNIME installation folder)