How is a data-driven approach better than random choice in label space division for multi-label classification?

Devs

How is a data-driven approach better than random choice in label space division for multi-label classification? | Read Paper on Bytez