Parsimonious clustering of time series

Time series arise in many areas, including engineering, computer science, medical science, social science and economics. Clustering of time series has become an important topic, motivated by the increased interest in these type of data. Several clustering algorithms have been developed to different type of time series. Most of the time, data must be preprocessed, i.e. by modeling each series with an appropriate model for time series. In this work, we propose a new clustering time series way which can be considered as belonging to both model-based and feature-based approach. The proposal consists of model each series by a penalized spline (P-spline) smoothers and then cluster the estimated spline coefficients. This method has a number of advantages, including excellent performance of resulting clustering procedure and a reduced computational time.