In order to solve the problem that how to explore and exploit the high-order label correlations effectively in multi-label learning

a Multi-Label evolutionary HyperNetwork based on label Correlations (MLHNC) is proposed in this paper. In MLHNC

the predicting results obtained from any multi-label learning method are utilized as input of the model

the high-order correlations among labels are represented and explored by hyperedges

and the final prediction is made by integrating the label correlation and feature information. The experimental results on six multi-label datasets compared with three state-of-the-art multi-label learning methods show that the MLHNC not only improves the performance of various state-of-the-art multi-label learning methods

but also provides readable learning results.