On-line Novelty Detection for Non-stationary Classes using a Modified Support Vector Machine

F. Camci and R.B. Chinnam (USA)


Novelty Detection, Support Vector Machine, Non stationary Data, Online Training


Novelty detection is the process of detecting abnormal behavior in a system by learning the normal behavior. Novelty detection has been of great interest to researchers from different domains, especially in areas where it is difficult or expensive to find examples of abnormal behavior (such as in machine diagnosis, medical diagnosis, and network attack detection). The literature offers several methods for novelty detection. Very few methods can handle novelty detection for non-stationary classes. To the best of our knowledge, there exists no method that can handle non-stationary data without making stringent assumptions about the class distribution. This paper proposes a novelty detection method, called Weighted Support Vector Novelty Detector (WSVND) for non-stationary classes using a modified Support Vector Machine. The paper also presents an on-line version of WSVND that is computationally efficient. Results from testing the proposed methods on three different datasets are very promising.

Important Links:

Go Back