Multi-label fuzzy similarity-based nearest-neighbour classification using association rule

Abstract

The demand for multi-label classification methods continues to grow in many modern applications, such as document classification, music categorisation, and semantic scene classification. This paper proposes two multi-label fuzzy similarity-based nearest-neighbour algorithms using the association rules. Specifically, in order to reduce the combination label number and avoid the label overlapping phenomenon, the association rule approach is employed to make the combination labels collapse to a set of sub-labels. Then by transforming the multi-label training data into the single-label representation data, the fuzzy similaritybased nearest-neighbour methods perform the classification label prediction. According to the extracted association rules, the resulting label set is the union of the predicted labels and their associated labels. Apparently, such result set will be more able to maintain the relevance between the labels. Empirical results suggest that the proposed approach can improve the performance and reduce the training time compared with other multi-label classification algorithms.