Cluster Analysis with K-Mean versus K-Medoid in Financial Performance Evaluation

Abstract

Nowadays there is a large amount of information at our disposal, which is increasing day by day, and right now the question is not whether we have a method to process it, but which method is most effective, faster and best. When processing large databases, with different data, the formation of homogeneous groups is recommended. This paper presents the financial performance of Hungarian and Romanian food retail companies by using two well-known cluster analyzing methods (K-Mean and K-Medoid) based on ROS (Return on Sales), ROA (Return on Assets) and ROE (Return on Equity) financial ratios. The research is based on two complete databases, including the financial statements for five years of all retail food companies from one Hungarian and one Romanian county. The hypothesis of the research is: in the case of large databases with variable quantitative data, cluster analysis is necessary in order to obtain accurate results and the method chosen can bring different results. It is justified to think carefully about choosing a method depending on the available data and the research aim. The aim of this study is to highlight the differences between the results of these two grouping procedures. Using the two methods we reached different results, which means a different evaluation of financial performance. The results demonstrate that the method chosen for grouping may influence the assessment of the financial performance of companies: the K-Mean method produces a greater variety of groups and the range of results obtained after grouping is larger; whereas, the group distribution and the results obtained by the K-Medoid method are more balanced.