A new label ordering method in Classifier Chains based on imprecise probabilities

•The Classifier Chains (CC) are simple and effective to exploit label correlations in Multi-Label Classification.•The label order strongly influences the performance of the CC.•We propose a new label ordering method that uses imprecise probabilities to estimate the label correlations.•The proposed procedure improves the label ordering methods based on label correlations proposed so far.•Our proposal obtains better experimental results than the label ordering methods developed so far based on label correlations.