» ГЛАВНАЯ > К содержанию номера
 » Все публикации автора

Журнал научных публикаций
«Наука через призму времени»

Декабрь, 2017 / Международный научный журнал
«Наука через призму времени» №9 2017

Автор: Носов Андрей Сергеевич, Студент Института экономики и менеджмента
Рубрика: Экономические науки
Название статьи: Using neural networks to evaluate the credit benefit of the borrower

Статья просмотрена: 141 раз

УДК 2964


Nosov Andrei


Kemerovo State University, Kemerovo

Abstract. In our time, the world is rapidly changing and new technologies are coming to our lives. In this article, we are trying to disassemble one of the newest technologies, namely the neural network, namely its application in order to assess the borrower's creditworthiness.

Keywords: neural network, credit rating, modern technologies.


In this article, I would like to raise the problem of using modern technologies in business and banking, in particular the use of such an innovative technology as a neural network. To date, the world is full of uncertainty and risk, the external environment of economic objects is very volatile and little predictable. First of all this is due to the rapid development of IT technologies that broke into our lives. Undoubtedly, the world has changed very much over the past ten to twenty years, thanks to the opportunities that IT technologies gave to business. Today, the lion's share of the processes is automated and does not require human participation, although recently we could not imagine that the place of the man behind the assembly line at the Ford plant will be occupied by a car. But now this is a reality the world of technologies is developing and this is pushing the economy forward.

And today in this article we will see that in the near future some professions and specialties that seem to us an integral part of society will not be needed. Consider this issue in the sphere of banking services, namely in the sphere of lending to individuals. As you know, the average size of a loan to an individual is much less than that of a legal entity. At the same time, the cost of issuing a loan to an individual is quite high, in particular, the analysis of a person's creditworthiness and the cost of wages to a bank employee reduce the profitability of the loan. Naturally, this raises the cost of this service. At the moment, banks are assessing the creditworthiness of individuals on the basis of a scoring model. This approach is based on the use of a quantitative risk assessment calculated by the credit institutions themselves based on the results of a comprehensive analysis of the borrower's activities. At the same time, the borrower's financial position, the quality of servicing the loan debt, the borrower's gender, the number of children, education, how many years he has been employed in his organization and other information available to the bank are taken into account. Scoring is designed to collect heterogeneous data, to give them weight. As a result, the client of the bank receives a scoring score and is referred to a certain group by the degree of risk of non-return of funds borrowed. For all its advantages, it can not be said that the scoring model of creditworthiness assessment is without flaw. It has a number of drawbacks: it needs to be reviewed at least once every six months; banks independently subjectively develop this model; it is necessary for a person to participate in a full evaluation. [3]

At this stage of development, there is a technology that can replace the scoring model in the foreseeable future, and oust specialists from the market, in assessing credit risk. This technology is a neural network. The neural network is a sequence of neurons connected by synapses. The structure of the neural network has come to the programming world straight from biology. Thanks to this structure, the machine is able to analyze and even memorize various information. Neural networks are also able not only to analyze incoming information, but also to reproduce it from their memory. A neuron is a computing unit that receives information, performs simple calculations on it, and passes it on. They are divided into three main types: input, hidden and output. There is also a displacement neuron and a contextual neuron. In the case where a neural network consists of a large number of neurons, the term layer is introduced. Accordingly, there is an input layer that receives information, n hidden layers (usually there are not more than 3 of them) that process it and an output layer that outputs the result. Each of the neurons has 2 basic parameters: input data (input data) and output data (output data). In the case of the input neuron: input = output. In the rest, the total information of all neurons from the previous layer gets into the input field, after which, it is normalized using the activation function and falls into the output field. [2]

But within the framework of this article, we will not delve into the work of neural networks, this is not our ultimate goal. It is important for us to understand that the neural network is trained, in addition it accumulates experience and makes fewer mistakes with each new stage of training. A new neural network can be compared to a child who, every year, improves and makes fewer mistakes, relying on his experience.

This gives us great opportunities in the field of lending. We can learn the neural network by introducing data that can not be taken into account in scoring analysis!

You can consider the introduction of a neural network using the example of abstract bank X. Assume that Bank X has information about its customers, as well as people who have been denied credit. We make create incoming neurons and teach the neural network based on this data. We will enter the parameters of the scoring model, and the decision of the bank to issue a loan. If there is a failure, we will have one output neuron. When issuing a loan, there may be several outcomes:

• The loan was returned on time;

• The loan is not returned;

• Delayed payments.

By inserting these data, we will teach the neural network to work similarly to the scoring model. But, with further training, the neural network is much better than the scoring model of evaluation, since it will determine the weights of the factors, establish interrelations, and therefore it will learn! The neural network will independently determine market trends and respond to them, much faster than a person.

In addition to these indisputable advantages of the neural network there are a few more, no less significant. We can withdraw lending through the bank's website to a new level! This will not just leave the application and a preliminary decision to grant a loan. It is possible to make such a level of credit rating on the basis of a neural network that the bank will be able to issue loans online with minimal risks. This can be achieved through modern analysis technologies. The neural network can check the social networks of the borrower, recognize the photographs (for example, if the borrower often happens abroad, then he has higher solvency), learn geolocation, client routes and draw conclusions about his solvency! Also, the bank will be able to reduce labor costs, as the evaluation staff will no longer be needed. We need a reduced staff to work with the neural network.

At the moment, there are already examples of the neural network being squeezed out by some employees in the banking sector. In Sberbank was dismissed four hundred and fifty lawyers. They were replacing the neural network. Herman Gref (CEO of Sberbank) advised lawyers who do not understand neural networks, master new technologies. As he said, with work on preparation of statements of claim neural networks now consult better people.[1]

As a result, I want to say that the world is constantly changing and our task is not to lag behind the current trends, to keep pace with the times. And these are not just words, it must be done, at least, in order to remain an actual labor force in the labor market and find a place in the professional sphere. Of course, it is difficult to predict the changes that will occur in a few years, let alone forecasts for a longer-term perspective, but even now one can say that the technology of neural networks will change the world, as we saw after examining its possible impact on banking.



Список литературы:

  1. Official website of RBC TV channel:// http://www.rbc.ru/business/23/07/2017/5974b7a69a79477896b6708d
  2. Portal about programming and modern technologies:// https://habrahabr.ru/post/312450/
  3. International convergence of capital measurement and capital standards: refined framework approaches (Basel2) / Bank for International Settlements, 2012 – P. 266


Фамилия Имя Отчество: