Multiple Regimes Model Reconstruction Using Symbolic Time Series Methods

Abstract

In this paper we describe and apply the methods of Symbolic Time Series Analysis to an experimental framework. We discuss data symbolization as a tool for identifying temporal patterns in experimental data and use symbol sequence statistics in a model strategy. In particular, we introduce a static partition in a time series of inflation rates. This partition is based on economic criteria using the notion of economic regime. Consequently, the time series is converted into a symbolic sequence. The probability of occurrence of different symbol strings constitutes the symbol sequence statistics. Then a method is discussed for reconstructing a model of inflation fluctuations from measured time series data, where the symbol sequence statistics are used as the target for reconstruction. That is, we will show how the observed symbolic sequence statistics can be used as a target for measuring the goodness of fit of the proposed model.