High-dimensional and wide-scale anomaly detection using enhancing support vector machine

For multidimensional data, difficult problems are encountered in the anomaly detection process. Unrelated features can hide the presence of anomalies in an indeterminate way. This multidimensional problem, is a serious obstacle to be overcome for many anomaly detection techniques. The creation of a robust anomaly detection model for multidimensional data requires a combination of an unsupervized feature extractor and an anomaly detector. Support vector machines are used efficiently when generating feature vectors, but they may be inefficient in modeling operations in multidimensional data sets. Multilayer neural networks stractures are one of the techniques frequently used to identify underlying attributes. In this paper, a extended support vector machine was used together with unsupervised multilayer neural networks and the results obtained in the extraction process of the self-efficiency, computation complexity and scalability. As a result of the study, the results of these tests are compared and reported.