String attributes get lost when using sparse instances

From BioWeka

Contents 1 Circumstances 2 Symptoms 3 Causes 4 Solutions

Circumstances

• Using XslXmlLoader on command line.
• The stylesheet outputs sparsed instances.
• The instances have one or more string attributes.

Symptoms

• The values of the string attributes of the first instance get lost.

Causes

• To output the instances in the sparsed ARFF format the ArffSaver class is used. As described on http://www.cs.waikato.ac.nz/~ml/weka/arff.html there is a problem with sparse instances and string attributes.

Solutions

• Edit the XML file and add a dummy instance as the first instance.