DOI: 10.32725/978-80-7394-976-1.33

Functional cluster regression for commodities and the representatives of stock indexes

Gabriela Hlásková, Tomáš Mrkvička

University of South Bohemia in České Budějovice, Faculty of Economics, Studentská 13,370 05 České Budějovice, Czech Republic, hlaskg00@ef.jcu.cz.

University of South Bohemia in České Budějovice, Faculty of Economics, Studentská 13, 370 05 České Budějovice, Czech Republic, mrkvicka@ef.jcu.cz.

While using external variables as potential predictors, one might be challenged by numerous possible variables, which while used once-on-time might devalue the predictive ability of individual ones. Thus, the pre-selection of relevant possible predictors should be used. For purpose of risk prediction of exchange rate changes, many external variables (time series) are available, thus most commonly traded ones were selected in the final number of 32 variables. Only a fraction of those should be in fact used while proceeding with the computation. Thus, out of many possible methods of selection of finding the relevant variables, the functional cluster analysis would be used. In this paper, we describe a case study of the functional cluster analysis application on time series as one of the possible methods of explanatory variable selection for the exchange rates.
 

Keywords: predictors, commodities, functional cluster analysis, stock indexes exchange rate

pages: 222-228