Feature Selection Algorithm for Multi-label Classification Based on Graph Operations

Referring to the method of Guo Yankui [1] and others in the field of single classification, this paper proposes a correlation attribute based on graph operation. Multi-label classification and selection algorithm. The algorithm takes the correlation between labels and attributes and between attributes as the weights of bipartite graphs and complete graphs respectively, sets thresholds for graph operation, constructs maximal connected subgraphs, and finally obtains the optimal attribute subset. This subset can effectively contain the information in the original data set. In this paper, the commonly used multi-label classification data sets are selected for experiments. Experiments show that this algorithm can effectively reduce the data dimension, reduce redundant information and improve the classification rate in multi-label data sets.