Overview of one-Class Classification

One-class classification, which was tested successfully in unbalanced sample classification problems, is one of the hotspots in pattern recognition research. This paper first analysis the commonly used one-class classification methods, then classifies these methods into three categories: boundary based method, re-construction based method and border based method. Finally, a series of testing experiments based on artificial database are designed to test the advantages and disadvantages of these methods from the aspect of learning ability, classification decision and algorithm complexity.