Evaluation of Criteria for the Classification of Enterprises

  • Vytautas Boguslauskas Kaunas University of Technology
  • Ruta Adlyte Kaunas University of Technology
Keywords: performance criteria, classification of enterprises, cluster analysis, Mahalanobis distance


Successful development and growth of enterprises is of great importance for the economic growth of the country, social stability and creation of new work places. However, the enterprise’s activity is directly or indirectly influenced by internal and external factors. An effective strategy of companies might be one of the most important factors. Loss-making enterprises can improve their results following the strategy of profitable enterprises, however the question is, which enterprises could be attached to the domain of profitable ones and what classification criteria should be applied. The analysis of special literature has shown that it is not enough to analyze separate activities of enterprises, but the whole system of variables should be analyzed. Cluster analysis is a statistical method, which allows classifying the selected objects into the classes according to their similarity within the class and significant difference between the classes. This classification is based on the analysis of all parameters of the system, so it could be effectively used for the grouping of enterprises into the classes. The main purpose of this investigation was to determine the classification criteria for enterprises according to their Net Profit. Profit (loss) and balance sheet data of 50 profitable and 50 loss-making enterprises during the year 2002-2006 were taken for the investigation from the Lithuanian Department of Statistics. After the first term analysis the financial data which were not characteristic for the majority of the analyzed enterprises were excluded from the investigation as well as the data with outliers. 30 parameters corresponding to the different profit (loss) and balance sheet lines were selected for the further analysis. Collinearity diagnostics of data applying three different methods was performed in the second step. Not correlated parameters were included into regression model. Three profit describing regression equations were composed, variables of which were used for the classification of the enterprises. Hierarchical clusterization methods were used for the classification of enterprises into profitable, loss-making and mixed (all others). Performing cluster analysis the selection of the linkage distance and the classification method was performed first of all. The best linkage distance and the best classification method is that one, according to which most of the profitable and loss-making enterprises match their classes. The investigation has shown that the Ward’s method and Euclidian distance were most suitable for the classification. Due to this reason, the enterprises were classified using different variables, which were included into the regression equations. Only independent variables of the regression analysis equation were used in the first case; the independent variables and net profit (dependent variable) were used in the second case and weighted variables of the regression analysis equation were used in the last case. The best result was achieved using independent variables selected by correlation analysis and included into regression analysis equation composed for the evaluation of net profit. In this case 17 profitable and 21 loss-making enterprises matched the classification criteria. Mahalanobis-Taguchi system was used for the evaluation of the differences between profitable and lossmaking enterprises. Present paper discusses the necessity to establish standard set of profitable enterprises and provides a validation of this standard. Performed investigation has shown that the classification of enterprises according to selected variables credibly evaluates performance of the enterprises and makes it possible to classify them into the classes of profitable and loss-making companies.