There is growing literature in macroeconomics, especially on business cycle synchronization, employing different methods of time series clustering. However, even as an unsupervised learning method, this technique requires making choices that are nontrivially influenced by the nature of the data involved. By extensively testing various possibilities, we arrive at a choice of a dissimilarity measure (compression-based dissimilarity measure, or CDM) which is particularly suitable for clustering macroeconomic variables. We check that the results are stable in time and consistent with the literature on core-periphery pattern of European business cycles. We also successfully apply our findings to the analysis of national economies, specifically to identifying their structural relations. To our knowledge, it is the first comprehensive analysis of the usefulness of the different dissimilarity measures for the macroeconomic research.
time series clustering similarity cluster analysis GDP
E00 C18 C63
A slightly updated version of this paper has been published in ECONOMETRICS. EKONOMETRIA. Advances in Applied Data Analysis, ISSN 1507-3866, e-ISSN 2449-9994, Vol. 22 (2018), No. 2, p. 74-88, DOI: https://doi.org/10.15611/eada.2018.2.06. A more recent version is available at the link below.