Detection of Anomalies and Novelties in Time Series with Self-Organizing Networks

This paper introduces the DANTE project (Detection of Anomalies and Novelties in Time sEries with self-organizing networks), the goal of which is to evaluate several self-organizing networks in the detection of anomalies/novelties in dynamic data patterns. In this paper, we first describe three standard clustering-based approaches which use well-known self-organizing neural architectures, such as the SOM and the Fuzzy ART algorithms, and then present a novel approach based on the Operator Map (OPM) network [1]. The OPM is a generalization of the SOM where neurons are regarded as temporal filters for dynamic patters. The OPM is used to build local adaptive filters for a given nonstationary time series. Nonparametric confidence intervals are then computed for the residuals of the local models and used as decision thresholds for detecting novelties/anomalies. Preliminary simulations suggest that the proposed approach consistently outperforms standard clustering-based algorithms.