System and method for extracting symbols from numeric time series for forecasting extreme events

摘要

A method for extracting symbols from a numeric time series comprises the steps of: receiving a finite time series of data elements for a particular application, the data elements characterized as having one or more sharp changes in values; for each sharp change in the finite time series, extracting a window of elements from the time series that precedes each sharp change; building a matrix from the time series window extracts; performing singular value decomposition on the built matrix to obtain characteristic matrices; and, obtaining vectors of symbols from resulting characteristic vectors determined from the singular value decomposition step, wherein the resulting symbols are used by forecasting algorithm to predict a future sharp change in subsequent finite time series received for the application. The sets of finite symbols obtained may be used as the basis for applications requiring the prediction of an extreme change in a received time series.