Non-Negative Matrix Factorization and Support Vector Data Description Based One Class Classification

One class classification is widely used in many applications. Only one target class is well characterized by instances in the training data in one class classification, and no instance is available for other non-target classes, or few instances are present and they cannot form statistically representative samples for the negative concept. A two-step paradigm employing nonnegative matrix factorization (NMF) and support vector data description (SVDD) for one class classification training of nonnegative data is developed. Firstly, a projected gradient based NMF method is used to find the hiding structure from the training instances and the training instances are projected into a new feature space. Secondly, SVDD is employed to perform one class classification training with the projected feature data. Classification examples demonstrate that the proposed method is superior to principal component analysis (PCA) based SVDD method and other standard one class classifiers.