I have an ARFF file with a column of 32-bit hash values of type string. I am trying to convert it from string to nominal using the filters in in the Weka GUI (select all then filters > unsupervised > StringToNominal) but it doesn't seem to have any effect and remains type string. What am I doing wrong?
2 Answers
Solved the problem in the end by converting the alphanumeric hashes into numeric hashes using the DJB2 Hashing Algorithm in Knime before importing the ARFF file into Weka. Then I was able to use the numeric to nominal filter to convert the hash to nominal (filters > unsupervised > NumericToNominal).
I left this answer for future references.
You must specify the range of columns to which apply the selected filter. On the Weka interface, next to the "choose" button, click over the filter name and a configuration window will appear. Then, on the range attribute you can specify the columns number to which the filter will be applied. In this:
image the filter will be applied to the sixth attribute (UC). This is also observed in the weka command, in the argument -R.