Abstract:
Data mining approach for detection and prediction of anomalies in
high dimensional data is an essential task in dynamically evolving
systems. Many resources have been invested in order to capture
system's behavior while realizing that significant information  is
often missing or hard to reveal in high-dimensional data. In this
talk, we apply recently introduced diffusion framework that
analyzes these problems. In particular, we detect anomalies on data
that was collected by a performance monitor. This system handles
huge number of transactions and it is not allowed to be down.