Multi-label learning deals with data associated with a set of labels simultaneously. Dimensionality reduction is an important but challenging task in multi-label learning. Feature selection is an efficient technique for dimensionality reduction to search an optimal feature subset preserving the most relevant information. In this paper, we propose an effective feature evaluation criterion for multi-label feature selection, called neighborhood relationship preserving score. This criterion is inspired by similarity preservation, which is widely used in single-label feature selection. It evaluates each feature subset by measuring its capability in preserving neighborhood relationship among samples. Unlike similarity preservation, we address the order of sample similarities which can well express the neighborhood relationship among samples, not just the pairwise sample similarity. With this criterion, we also design one ranking algorithm and one greedy algorithm for feature selection problem. The proposed algorithms are validated in six publicly available data sets from machine learning repository. Experimental results demonstrate their superiorities over the compared state-of-the-art methods.
