A context-based approach for partitioning big data