ARFF reader/writer for MATLAB is a small library that support Weka's ARFF files without using any native Java binding.
It comes with a couple of usage examples:
- ARFF file writing
- ARFF file reading
Keep in mind that some features are missing and the ARFF spec isn't completely covered at this time. However for common operations (like parsing numeric, real, integer, string, date and last but not least nominal specification datatypes) and not-mission-critical use the whole lib is fine.
Feel free to extend the support or add the missing features by cloning the git repository and sending a pull request.
% arff_file => output file name (.arff or .arff.gz for compressed format)
% data => struct array (size x attributes) of dataset data
% relname => dataset relation name
%
% nomspec => struct of nominal specification mappings (optional)
% nomspec.field_class = { 'one', 'two', 'three' };
arff_write(arff_file, data, relname, nomspec);If using the .arff.gz extension keep in mind that the compression is done by
saving a .arff file to a temp dir (using OS default preferences), outputing its
compressed version using the arff_file path and finally removing the temp file.
% arff_file => output file name (.arff or .arff.gz for compressed format)
%
% data => struct array (size x attributes) of dataset data
% relname => dataset relation name
% nomspec => struct of nominal specification mappings (optional)
% nomspec.field_class = { 'one', 'two', 'three' };
[data, relname, nomspec] = arff_read(arff_file);If using the .arff.gz extension keep in mind that the decompression is done by
saving a .arff file to a temp dir (using OS default preferences), parsing its
content and finally removing the temp file.
For convenience is included a dummy dataset example_dataset.arff which
is used by the example scripts to create a couple of visualizations (reading the
data series and the nominal specification attribute):
