Nonnegative matrix factorization based one class classification

One class classification which has only one well-characterized target class in the training data is widely used in many applications. In fact the data set is often nonnegative in many real-life applications. Nonnegative matrix factorization (NMF) is able to find the hiding structure and information from the nonnegative instance data. NMF based one class classification method is presented and compared with other common one class classifiers and principle component analysis based one. A few classification examples demonstrate that NMF based method is efficient to improve the performance of one class classifiers those include density estimation method, reconstruction method, and boundary method. And the proposed NMF based method is superior to principle component analysis based one class classifiers.