Representation and Similarity Evaluation of Symbolic Time Series Data in Uncertain Environments

Coping with time series cases is becoming an important issue in employing case based reasoning (CBR) in many industrial and medical applications. This paper studies the representation and similarity matching of symbolic time series data that are generated from an uncertain (nondeterministic) process over time. We propose a method to convert such (lengthy) time series into a concise form to capture the stochastic and dynamic property of the underlying process. We also suggest several ways to build a similarity model based on the proposed scheme for representation of symbolic time series cases. Hopefully the work in this paper could offer an initial step to contribute to a probabilistic CBR framework tackling both temporal factor and stochastic uncertainty in a unified manner.