Scholarly record

METRIC METHODS OF DATA MINING IN THE ASSESSMENT OF ENVIRONMENTAL AND ECONOMIC RISKS OF BUSINESSES

Igor V. Arinichev

Abstract

In the article the authors propose to use inductive metric data mining methods to assess the environmental and economic risks of businesses in various fields of activity and industry. The purpose of this methodology is to conduct operational monitoring of both environmental and economic condition of enterprises to regularly monitor the dynamics of internal and external factors when planning strategic decisions. A distinctive feature of the proposed approach is the integration of environmental and economic indicators into the system of qualitative indexes presented by expert assessments of six types of risks: industry risk, management risk, financial flexibility, environmental risk, competitiveness and operational risk. At the same time, each index is evaluated by experts on three levels: ?Positive?, ?Average? and ?Negative?. In the article, the authors conduct a comparative analysis of five groups of methods with different types of distance functions, including the Euclidean metric and Chebyshev distance, as well as four types of kernels for the method of potential functions. A quality cross-check is carried out using training and test samples in order to configure the parameters of the algorithms. The simulation results showed that for some of the metrics in the feature space, the considered training methods are in good agreement with the initial data and demonstrate a slight error on the test data. Training and optimization of the algorithms were carried out in the development environment Visual Studio 2017 using C # programming language. Implementations of metric methods, their reliability, the ability of algorithms to analyze significant amounts of information is relatively simple. Given this, the authors suggest that the proposed approach is of practical value, providing a comprehensive basis for making management decisions.

Publication Impact Profile

• Captures
• Mendeley - Readers: 7

PLUMX see details

DOI resolver: doi.org/10.5593/sgem2019/5.3/s21.065

Publication details

Back to publication list