This document discusses techniques for anomaly detection in time series data using symbolic representations. It describes mapping time series windows to symbols in a finite symbol space using SAX (Symbolic Aggregate approXimation) or self-organizing maps. A metric like log-likelihood ratio is then used to compute how anomalous a symbol is based on its observed frequency. These techniques can predict events by identifying correlations between symbolic representations and future outcomes. An example of predicting hard drive failures using SAX and sensor data is provided.