Combining Functional Data Projections for Time Series Classification

摘要

We afford the classification of time series in the Functional Data Analysis (FDA) context. To this aim we introduce projections methods for the time series onto appropriate Reproducing Kernel Hilbert Spaces (RKHSs) with the aid of Regularization Theory. Next we project the curves onto a set of different RKHSs. Then we consider the induced Euclidean metrics in these spaces and combine them in order to obtain a single kernel valid for classification purposes. The methodology is tested on some real and simulated classification examples.