Functional Cluster Analysis of Financial Time Series

In this paper we address the problem of clustering functional data. In our applications functional data are continuous trajectories evolving over time. Our proposal is to cluster these trajectories according to their sequence of local extrema (maxima or minima). For this purpose, we suggest a new dissimilarity measure for functional data. We apply our clustering technique to the trajectories of the shares composing the MIB30 stock index computed at the Milan Stock Exchange Market, paralleling the contribution of Ingrassia and Costanzo in this Volume.