Financial Time Series Classification by Nonparametric Trend Estimation

This work considers the classification of financial nonstationary time series, where the nonstationarity is due to the presence of a deterministic trend. It is evaluated in a high-dimensional context by looking at the first derivative of the trend function and without requiring a pre-specified form. This is achieved by means of a nonparametric estimator which is used in a two stage procedure: the first stage selects the time series with no trend and the second stage focuses the attention on nonlinear trends. A real data application to US Mutual Funds is conducted to demonstrate the validity and applicability of the procedure.