Time series classification using adaptive dynamic targets

To train a classifier with supervised learning appropriate targets have to be provided. In the case of time series this can be complicated if there is only one target for the whole time series, but the learning algorithm needs a target at each time step. In the paper a technique is introduced, which is able to provide appropriate targets at each time step. As a result of this technique the impact on classification of each time step is determined, which is very useful in applying the trained classifier to new data. The paper describes this technique in detail and the basic findings of experiments on artificial data and real world data are given.