Abstract

Several approaches to detect or even predict abnormal events as early as possible will be discussed. The model input is a time series of frequently collected data. The approaches presented in this document use various methods originating in the field of data mining, machine learning and soft computing in a hybrid manner. After a basic introduction including several areas of application, the focus will lie on the modular parts of the proposed server outage model, starting with a discussion about different approaches to time series prediction such as SARIMA models and specific artificial neural networks. After the presentation of several algorithms for outlier detection (angle-based outlier factor, one-class support vector machines) the gained results of the simulation are put up for discussion. The text ends with an outlook for possible future work.