Font Size: 

In this article, we propose a new approach to the classification of multivariate time series. We use a functional approach to data analysis and combine information from raw data and functional derivatives. To provide a comprehensive comparison, we conducted a set of experiments, testing effectiveness on fifteen multivariate time series datasets from a wide variety of application domains. Our experiments show that this new method provides a more accurate classification of the examined datasets.

functional data, classification, discriminant coordinates, curvature