Classification for Time Series Data. An Unsupervised Approach Based on Reduction of Dimensionality

In this work we use a novel methodology for the classification of time series data, through a natural, unsupervised data learning process. This strategy is based on the sequential use of Multiple Factor Analysis and an ascending Hierarchical Classification Analysis. These two exploratory techniques complement each other and allow for a clustering of the series based on their time paths and on the reduction of the original dimensionality of the data. The extensive set of graphic and numerical tools available for both methods leads to an exhaustive and rigorous visual and metric analysis of the different trajectories, including their differences and similarities, which will turn out to be responsible of the classes ultimately obtained. An application from Finance, used previously in the literature, highlights the versatility and suitability of this approach.