.png){kind=icon}
.png){kind=icon}
Return
A one-class feature extraction method based on space decomposition
.png){kind=icon}
PRE
AI
DOI：10.1007/s00500-022-07067-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
One-class classification is an important branch of machine learning. Feature extraction is an important means to improve the performance of one-class classifiers, but there is no generalized method yet reported to solve this problem. In this paper, a framework is proposed for one-class feature extraction. The proposed framework divides the original feature space into two orthogonal spaces, namely the principal space and the complementary space. The principal space is used to learn the features of the target class, and the complementary space is used to learn the features of the abnormal class. The features extracted from the two spaces are fused as the final one-class feature vector of the original feature space. Furthermore, a specific implementation method, complete principal component analysis (CPCA), is proposed. First, CPCA conducts principal component analysis to calculate the projection scores of the target class samples in the principal space. Then, according to the projection vectors of the principal components (obtained in the principal space), the corresponding complementary space is constructed. The projection of the sample in the complementary space is calculated and transformed into the first-order norm as the extracted feature in the complementary space. Several datasets are used to verify the effect of this proposed method. The experimental results show that the proposed CPCA has good universality for one-class feature extraction problems.
Keywords:
One-class classification
Feature extraction
Principal space
Complementary space
Principal component analysis