Project CEMAPRE internal
| Title | Fragmented-periodogram and fragmented-autocorrelation approaches for clustering big data time series |
| Participants | Andreia Albino, Jorge Caiado (Principal Investigator), Nuno Crato |
| Summary | We will formally propose and study an autocorrelation procedure designed to characterize and compare large sets of long time series. This time-domain procedure is contrasted with a frequency domain approach recently introduced and discussed in the literature. In both procedures, instead of using all the information available from data, which would be computationally very expensive, adequate regularization rules select and summarize the most relevant information suitable for clustering purposes. Essentially, we suggest using the autocorrelation function of the time series computed only around the lags of greatest interest. Then, we will study this method in a few ways. First, we will argue theoretically that fragmenting the autocorrelation function can have efficiency advantages when comparing time series. Second, by means of a large simulation study, we will explore how the suggested procedure is able to condense the relevant information of the time series. We will then compare its results with those from the frequency domain counterpart. Third, we will illustrate this procedure in a study of the evolution of several stock markets indices. We further show the effect of recent financial crises over these indices’ behaviour. |