A computational study on 10 benchmark datasets found that MFLRA outperformed eight comparison methods by combining multi-order relevance, label uncertainty and feature interactions.