Cost-sensitive feature selection is an important research issue in both machine learning and data mining. Most existing cost-sensitive feature selection work deal with the single-label data. However, in real applications, the data usually is multi-label, continuous and incomplete because of the technology or cost limitations during data collection. To alleviate this problem, a cost-sensitive feature selection algorithm is proposed here for incomplete neighborhood multi-label which can implement feature selection based on considering about the weighted test cost. The experimental results show that our algorithm can select a low-cost feature subset without losing the classification accuracy. The effectiveness and feasibility of the proposed algorithm is verified by the performance on the three Mulan datasets.
