Multi-label data classification differs from traditional single-label data classification, in which each input sample participated with just one class tag. As a result of the presence of multiple class tags, the learning process is affected, and single-label classification can no longer be used. Methods for changing this problem have been developed. By using these methods, one can run the usual classifier classes on the data. Multi-label classification algorithms are used in a variety of fields, including text classification and semantic image annotation. A novel multi-label classification method based on deep learning and feature selection is presented in this paper with specific meta-label-specific features. The results of experiments on different multi-label datasets demonstrate that the proposed method is more efficient than previous methods.