An Improved Progressive Transductive Algorithm Based on One Class SVM

Based on the analysis of the One-Class SVM theory, we apply progressive transductive learning method to the One-Class SVM. During the process the training samples be progressively given by label and can be dynamically adjusted. This paper propose progressive transductive OCSVM learning methods. The PTOCSVM classifier can be used in the training process and effectively step by step to make identification and classification, It has self-repair of error and get better test results than directly transductive learning.