Error message

[Leonardo Lion]

Hi all, 

I'm doing text classification using MOA. I loaded my dataset into MOA which has bout 400,000 number of instances. All the instances are string ones.However, in order to perform classification, I used "NiveBayes". As a result, MOA said the process is failed, and produced this error message: 

Failure reason: weka/core/scripting/JythonSerializableObject

*** STACK TRACE ***java.lang.NoClassDefFoundError: weka/core/scripting/JythonSerializableObject

                at weka.core.SerializedObject.<init>(SerializedObject.java:94)

                at weka.core.Attribute.addStringValue(Attribute.java:933)

                at weka.core.converters.ArffLoader$ArffReader.getInstanceFull(ArffLoader.java:616)

                at weka.core.converters.ArffLoader$ArffReader.getInstance(ArffLoader.java:441)

                at weka.core.converters.ArffLoader$ArffReader.readInstance(ArffLoader.java:409)

                at weka.core.Instances.readInstance(Instances.java:984)

                at moa.streams.ArffFileStream.readNextInstanceFromFile(ArffFileStream.java:150)

                at moa.streams.ArffFileStream.nextInstance(ArffFileStream.java:114)

                at moa.tasks.EvaluatePrequential.doMainTask(EvaluatePrequential.java:190)

                at moa.tasks.MainTask.doTaskImpl(MainTask.java:50)

                at moa.tasks.AbstractTask.doTask(AbstractTask.java:57)

                at moa.tasks.TaskThread.run(TaskThread.java:76)

Why this error, and how to fix it?

Note that, MOA works perfectly with similar dataset that even has bigger number of string instances. 

Highly appreciate your advice. 

Leonardo

[Heitor Murilo Gomes]

Hi Leonardo, 

Were you able to load that same dataset on weka? 

Best Regards,

Heitor

[Leonardo Lion]

Hi Heitor, 

Yes, I can load it into Weka perfectly. 

Any idea about how to solve the error?

Thanks.

Leonardo

--
You received this message because you are subscribed to the Google Groups "MOA users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moa-users+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Albert Bifet]

>> java.lang.NoClassDefFoundError: weka/core/scripting/JythonSerializableObject
>> at weka.core.Attribute.addStringValue(Attribute.java:933)

Which version of MOA are you using? What is the format of the dataset?

Cheers, Albert

[Leonardo Lion]

I'm using 2014.11 release which I had it long time ago.

Leonardo

[Albert Bifet]

Does the header of your dataset contain string attributes? MOA does
not work with string attributes.

Cheers, Albert

[Leonardo Lion]

I'm doing text classification, hence, the dataset file has "@attribute text string" additional to the class attribute. Is there wrong with this sort of data?

Leonardo

[Albert Bifet]

Weka can process your file but MOA can not process string attributes.
You need to pre-process the file with WEKA filters to convert it to a
dataset with discrete or numeric values.

Cheers, Albert

[Leonardo Lion]

Thanks for helpful solution. However, it's difficult to Weka to handle big data (Unlike MOA) as it has problem with memory. In your opinion, how to handle big data using Weka without coding in order to avoid any heap problem?

Any suggestions would be greatly appreciated.

Leonardo

[Albert Bifet]

With 400,000 instances, you should be able to use the
StringToWordVector filter in Weka and then process the resulting
dataset with MOA.

Cheers, Albert

[Leonardo Lion]

Perfect, thank you so much for this very helpful advice. 

Regards, 

Leonardo