将输入arff文件拆分为较小的块以处理非常大的数据集

I am trying to run a weka classifier on map reduce and loading entire arff file of even 200mb is leading to heap space error, so I want to split the arff file into chunks, but the thing is it has to maintain the block information ie the arff attributes information in every chunk so as to run the classifier in each mapper. Here is the code that I am trying to split the data but not able to do with efficiency,I am trying to run a weka classifier on map red