The application of data mining to explore the connections between the quality of organization of companies and their business results

Marko Pregeljc (2011) The application of data mining to explore the connections between the quality of organization of companies and their business results. PhD thesis.

PDF Download (2626Kb)

Abstract

The thesis deals with searching and proving the connections between the quality of organization and levels of economic outcomes of formal social units (firms). Therefore, it first theoretically defines and distinguishes the basic concepts of organizations and formal social units (firms) according to Lipovec’s developed theory of organization and its further development regarding the Mihelčič’s identification and segmentation of critical target organizational relationships. From there, the original Slovenian method of assessing the quality of organization of formal social unit MUKOZ is derived, which is used as a basis for assessing the organizational impacts on the economic results of the firm. To assess these effects, approach of data mining is used and initially theoretically briefly presented, from the regression, classification, clustering and attribute quality measures. The latest general method for explanation of regressors and classifiers is also used, which is independent of the applied methods of machine learning and the experiments raise expectations for very good results. Also, the thesis presents the approach of text mining, from the essential differences between text and numeric data, transformations between them, creating words in a dictionary, stemming to a basic form as well as classifying and clustering texts and its performance assessment, all in functional connection with voluminous text of the method for assessing the quality of organization. On the collected data from the formal social units in Slovenia the links between organizational indicators and economic outcomes are presented, separately for the economy and for the public sector. The findings of classical data mining methods and new general method for model's interpretation are cross compared, both confronted with standard statistic approach of Spearman's rank correlations in order to substantiate the links between indicators of the quality of organization and appropriate economic outcomes. On the model of assessment of the quality of organization MUKOZ the comparative analysis of classification algorithms for organizational problems is accomplished in order to verify the consistency, reliability and validity of this model. With the same purpose on the extensive questionnaire of this method text analysis is performed to further confirm its logical structure. On the elements of this model, three different clustering approaches are applied, based on the frequency of words, the binary presence / absence of words and TF-IDF measure, to prove throug data mining methods the substantial congruity with the current clustering model. At the same time, some new concepts are exposed to be considered in the future development of this method.

Actions (login required)