Self-advised Incremental One-Class Support Vector Machines: An Application in Structural Health Monitoring

Ali Anaissi, Nguyen Lu Dang Khoa, Thierry Rakotoarivelo, Mehri Makki Alamdari, Yang Wang

International Conference on Neural Information Processing

Incremental One-Class Support Vector Machine (OCSVM) methods provide critical advantages in practical applications, as they are able to capture variations of the positive samples over time. This paper proposes a novel self-advised incremental OCSVM algorithm, which decides whether an incremental step is required to update its model or not. As opposed to existing method, this novel online algorithm does not rely on any fixed threshold, but it uses the slack variables in the OCSVM as proxies for data in order to determine which new data points should be included in the training set and trigger an update of the model’s coefficients. This new online OCSVM algorithm was extensively evaluated using real data from Structural Health Monitoring (SHM) case studies. These results showed that this new online method provided significant improvements in classification error rates, was able to assimilate the changes in the positive data distribution over the time, and maintained a high damage detection accuracy in these SHM cases.

Publication Type


Publication Date