Learning target class eigen subspace (LTC-ES) via eigen knowledge grid

In one-class classification (OCC) tasks, only the target class (class-of-interest (CoI)) samples are well defined during training, whereas other totally absent. OCC algorithms, high dimensional data adds computational overhead apart from its intrinsic property of curse dimensionality. For learning, conventional dimensionality reduction (DR) techniques not suitable due to negligence unique statistical properties CoI samples. this context, present research proposes a novel guided DR technique extract eigen knowledge grid that contains most promising eigenvectors variance-covariance matrix process lower and higher eigenvalued rejected via analysis because variance may split itself, do contribute significant information. Furthermore, identified is utilized transform subspace. The proposed approach named as learning subspace (LTS-ES) ensures strong separation classes. To show effectiveness transformed subspace, one support vector machine (OCSVM) has been experimented on wide variety benchmark datasets in presence of: original feature space, features obtained approximately 80%-90% cumulative variance, 50% variance. Finally, new performance measure parameter called stability factor introduced validate robustness approach.