A One-class Classification Algorithm Based on Steiner Minimal Tree of Typical Samples Covering Model in High-dimensional Space

Minimum Spanning Tree Class Descriptor (MSTCD) describes the target class with the assumption that all the edges of the graph are also basic elements of the classifier which offers additional virtual training data for better description of sample distribution in high dimensional space. However, this descriptive model has too many branches, which makes the model more complicated, and its local coverage is not so reasonable. In this case, according to the continuity law of the feature space of similar samples, a one-class classification algorithm based on Steiner minimal tree of typical samples covering model is presented in this paper. The method first prunes the training set, eliminates redundant information and noise information and selects the most representative samples as a new training set; then it builds Steiner minimal tree covering model on the retained typical samples. Theoretical analysis and simulation experimental results show that the presented method can describe the distribution of target class more reasonably, construct more reasonable covering model without increasing the model complexity. It performs better than MSTCD in accuracy of classification and applicable sample size.