How many attributes in one arff file that python-weka-wrapper can support when clustering?

Question

I want to know if there's any limitations of dataset for python-weka-wrapper.

I'm doing the ClusterEvaluation(with weka.clusterers.HierarchicalClusterer) via python-weka-wrapper 0.3.8. And I can get the final clustering results with smaller dataset (62 attributes) successfully. But with a larger dataset (264 attributes), it stuck about 2 days without any output debugging message with python-weka-wrapper. Using the same arff(264 attributes), it can run correctly with WEKA.

I want to know if there's any limitation of dataset for python-weka-wrapper. Or is there any setting I should check first for larger dataset?

Thanks!

Answer 1

python-weka-wrapper has the same limitations as Weka, as it just wraps around a Weka process running inside a JVM. However, you might have to increase the heap size used for the JVM. Use the max_heap_size parameter to set a different memory limit when starting the JVM (see documentation).