Entropy-based fuzzy clustering of interval-valued time series

This paper proposes a fuzzy C-medoids-based clustering method with entropy regularization to solve the issue of grouping complex data as interval-valued time series. The dual nature of the data, that are both time-varying and interval-valued, needs to be considered and embedded into clustering techniques. In this work, a new dissimilarity measure, based on Dynamic TimeWarping, is proposed. The performance of the new clustering procedure is evaluated through a simulation study and an application to financial time series.