DataSink

DataSink

The easiest way to save an weka.core.Instances object to a file is by using the weka.core.converters.ConverterUtils.DataSink class.

 import weka.core.converters.ConverterUtils.DataSink;
 import weka.core.Instances;

 Instances dataset = ...
 String outputFilename = ...
 try {
   DataSink.write(outputFilename, dataset);
 }
 catch (Exception e) {
   System.err.println("Failed to save data to: " + outputFilename);
   e.printStackTrace();
 }

Converter

You can use the ArffSaver class (weka.core.converters.ArffSaver) for saving a weka.core.Instances object to a file.

Here is the snippet :

 Instances dataSet = ...
 ArffSaver saver = new ArffSaver();
 saver.setInstances(dataSet);
 saver.setFile(new File("./data/test.arff"));
 saver.writeBatch();

Notes:

• using the converter approach, one can easily swap the ArffSaver with another saver, e.g., the CSVSaver to output the data in a different format.
• The Weka Examples collection dedicates quite a few examples to the use of converters in the wekaexamples.core.converters package (stable, developer)

Java I/O

You can also save the weka.core.Instances object directly using Java I/O classes:

 import java.io.BufferedWriter;
 import java.io.FileWriter;
 ...
 Instances dataSet = ...
 BufferedWriter writer = new BufferedWriter(new FileWriter("./data/test.arff"));
 writer.write(dataSet.toString());
 writer.flush();
 writer.close();

Note: using the toString() of the weka.core.Instances doesn't scale very well for large datasets, since the complete string has to fit into memory. It is best to use a converter, as described in the previous section, which uses an incremental approach for writing the dataset to disk.