Economic Performance Competitor Benchmarking using Data-Mining Techniques

In this paper we analyze comparatively the macroeconomic performance of different Central and Eastern European countries by the means of Data Mining (DM) techniques. We analyze the economic situations of three EU countries (Poland, Slovenia and Latvia), a newlyaccepted one (Romania), and other two non-EU countries (Russia and Ukraine). We have depicted economic performance of countries using a number of macroeconomic variables for the time period from 1993 till 2000. The economic variables that we used were: Currency Value (CV) (the inverse of the Exchange Rate ER), Domestic Prime Rate (Refinancing Rate – RR), Industrial Output (IO) compared to previous periods in percentages, Unemployment Rate (UR), Foreign Trade (FT) in millions of USA dollars. The dataset consists of monthly/annual data during the period 1993-2000, in total 225 cases with five variables each. As DM techniques we firstly applied Self-Organizing Map (SOM) to group the countries according to their economic performance. We applied a so-called “two-step” SOM clustering: firstly, we built larger maps that contained “raw” clusters and then, we re-grouped the “raw” clusters to form a smaller number of “real” clusters. We characterize each “real” cluster, using for each variable the following linguistic terms: VL – very low, L – low, A – average, H – high, VH – very high. Secondly, we go one step further and use the clustering results (of SOM) to build hybrid classification models to help position new countries’ performance within the existent economic performance clusters. This type of analysis can benefit the countries involved, EU in its monitoring process, business players such as international companies that want to expand their business and individual investors.