A hybrid intelligent classifier for anomaly detection.

The present research is focused on the use of intelligent techniques to perform anomaly detection. This task represents a special concern in complex systems that operate in different regimes. Then, this work proposes a hybrid intelligent classifier based on one-class techniques, capable of detecting anomalies of the different operating ranges. The proposal is implemented over an industrial plant designed to control the water level in a tank, taking into consideration three different operating points. The hybrid classifier is validated by using real anomalies, obtaining successful results.