Marketing research of credit rating on the basis of artificial neural network
Journal Title: Маркетинг і цифрові технології - Year 2017, Vol 1, Issue 2
Abstract
The aim of the article. The aim of this paper is to study the theoretical and methodological principles and development the practical guidelines for the assessment of credit rating of an enterprise by the help of neural networks. The results of the analysis. It is determined that the neural network is a dynamic system with the set of interconnected elementary processes and it is able to generate the output information in response to the input action. Artificial neural networks are the mathematical models as well as software or hardware applications, which are built on the principle of the organization and operation of biological neural networks such as networks of nerve cells of a living organism. We proved that in terms of machine learning the neural network is a special case of pattern recognition techniques, discriminant analysis, clustering methods. From a mathematical point of view, the training of neural networks is a multi-parameter problem of nonlinear optimization. A multi-layered neural network with a direct link is proposed to assess the credit rating of an enterprise. It was shown that processing of questionnaires and the decision of the classification problem was carried out with the help of "Neural Networks" package of the STATISTICA 8 Neural Networks (SNN) program, which is intended for clustering of potential bank customers. At the first stage of operation we proposed to prepare the output data; to carry out preliminary data processing for the input of the neural network should be realized at the second stage, and to form a neural network based on available data and carry out its training should be made at the third stage. To form the credit rating of the company we chosen MLP 21-13-1 model, which architecture is called multilayer perceptron. Multilayer Rosenblatt perceptron is a perceptron with additional layers of associative elements located between the corresponding sensitive (sensory) elements and reacting elements. We used the structure of the perceptron for the method of forming the image of the bank client-borrower, which characterized by the complete neurons connection, a double bond in which neurons of the first layer are associated with neurons of the second, which, in turn, have a connection with the neurons of the first layer. We proved that the comparison of the results of the conducted neural network analysis and the actual result of repayment of loans and debt servicing by the same sample of bank customers allows us to conclude that it is possible to further use the neural network analysis based on the multilayered perceptron, taking into account the assessment of the borrower. It is determined that the sensitivity analysis allows us to estimate the degree of influence of each factor of the questionnaire on the result: the smaller the rank which corresponds to variable, the greater its effect on the original parameter. Conclusions and direction for further research. As a result of the experiment conducted on the available sample, the neural network is constructed with an accuracy of 98.7%, which is a satisfactory indicator of its operation (the minimum number of correct predictions should be at least 80%).This allows us to conclude that usefulness of this software product to the banks for the automation of the assessment of potential customers' credit worthiness is appropriate. For a real offer, credit organizations need to replenish the client base, train the neural network for the further operation. Consequently, neural networks can be used as a reliable and effective tool for analyzing and forecasting socio-economic phenomena, including in the area of credit risk calculation for borrowing companies. It is advisable to make managerial decisions to increase the level of creditworthiness of enterprises by constructing an economic-mathematical model based on weight coefficients.
Authors and Affiliations
Sergii Smerichevskyi, Natalya Kasianova
Keywords
Related Articles
- EP ID EP249445
- DOI 10.15276/mdt.1.2.2017.2
- Views 138
- Downloads 0
Crowdfunding as a new challenge for doing business
This paper aims at clarifying the process of crowdfunding as an alternative source of financing for small and medium businesses. Its identifies the definition of crowdfunding and distinguished its four main categories, l...
MARKETING INTELLECTUAL PROPERTY
Marketing is variously defined. However, it is most often perceived as a process described in the Marketing-mix 4P, Marketing-mix 5p and Marketing-mix 7P and created values for consumers. The new marketing-mix concept de...
Determination of Logistic Risks in Seasonal Forecasts of Sales of mechanical engineering enterprises
The aim of the article. The data on which the researcher provides forecasts can have a different nature. It can have a steady or hopping dynamics. The data can be characterized by the current tendency to rise or fall, an...
Complexity of marketing and logistics strategies in the supply chain of fast moving consumer goods
The aim of the article. The purpose of the article is to justify the directions and conditions for achieving complementarity of marketing and logistics strategies in the supply chain of fast-moving consumer goods (FMCG)....
Strategy and tactics of internet marketing of distributors on the B2B market
The aim of the article. The purpose of the article is to investigate the features of Internet marketing in the B2B segment on the example of the field of flexible transport packaging, to identify the main strategic and t...