An analysis of time series representation methods
Vineetha Bettaiah, Heggere S. Ranganath · 2014
Because of high dimensionality, proven data mining and pattern recognition methods are not suitable for processing time series data. As a result, several time series representations capable of achieving significant reduction in dimensionality without losing important features have been developed. Each representation has its own advantages and disadvantages. In this paper, based on the requirements of key data mining applications, such as clustering, classification and query by content, characteristics desired in an ideal time series representation are identified. Using the identified characteristics as metrics, widely known time series representation methods are evaluated to determine the extent to which the representations satisfy the requirements.