Abstract

Recently, the large number of labels challenges the traditional multi-label classification methods in various application scenes. We introduce a framework of multi-label classification based on Boolean matrix decomposition for giving various multi-label classifiers the ability to predict labels in a high dimension label space. The BMD module in this framework satisfies two conditions of ‘the column use condition’ and exacts BMD, which can reduce the burden on training and predicting task of multi-label classifier to some extent, and the predicted result of multi-label classifier can be restored to original label space by simply Boolean multiplying with matrix. Experimental results on yeast datasets demonstrate that our framework can work particularly well on datasets with a large number of labels and obtain a better predicting accuracy. In summary, the methods discussed in this paper constitute important basic for utilizing more multi-label classifiers in high label dimension space, which is the main contribution of this paper.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call