UCI Blog Feedback
DS5

Dataset link:
https://archive.ics.uci.edu/dataset/304/blogfeedback 

Number of instances in training set:
60,021


Number of features:
280

Imbalanced ratio:
~30,000:1

This dataset already has a separate test set. so no need to split. However that test set is split into multiple files. 
so we need to combine them into a single .csv file.
The training set would have to be split into 4 folds for cross-validation.