On the Stratification of Multi-label Data

Author: Sechidis et al, 2011.

• Stratified sampling is a sampling method that takes into account the existence of disjoint groups within a population and produces samples where the proportion of these groups is maintained.
• Random distribution of multi-label training examples into subsets suffers from the following practical problem
  • it can lead to test subsets lacking even just one positive example of a rare label
• Two interpretations of multi-label stratification
  • Based on the distinct labelsets that are present in the dataset
  • Considers each label independently of the rest