Finance and Banking Developments (Banking and Banking Developments) by Charles V. Karsone

Chapter 1
CREDIT RATING MODELLING
BY NEURAL NETWORKS
Petr Hájek
Institute of System Engineering and Informatics, Faculty of Economics and
Administration, University of Pardubice, Studentská 84,
532 10 Pardubice, Czech Republic
Abstract
The chapter presents the modelling possibilities of neural networks on a complex real-world
problem, i.e. credit rating process modelling. First, current approaches in credit rating
modelling are introduced. Second, previous studies on corporate and municipal credit rating
modelling are analyzed. Based on this analysis, the model is designed to classify US
companies and municipalities into credit rating classes. The model includes data pre-

processing, the selection process of input variables, and the design of various neural networks’
structures for classification. The selection of input variables is realized using genetic
algorithms. The objective of this process is to select only significant variables in order to
improve the performance of neural networks. Input variables are extracted from financial
statements and capital markets in line with previous studies. These variables represent the
inputs of neural networks, while the rating classes stand for the outputs. The credit rating
classes have been obtained from the rating agencies Standard & Poor’s and Moody’s. Except
for exact credit rating classes, data are also labelled by investment or non-investment grades.
As a result, the classification accuracies and the contributions of input variables are studied
for the different number of classes. The results show that the rating classes assigned to bond
issuers can be classified with a high accuracy rate using a limited subset of input variables.
Keywords: Credit rating analysis, corporate credit rating, municipal credit rating, neural
networks, support vector machines, classification.
 1. Introduction
Credit rating is an independent evaluation whose aim is to find out how an object is
capable and willing to meet its payable obligations, specifically based on complex analysis of
all the known risk factors of the assessed object. The capability and willingness to meet
obligations is called creditworthiness. More precisely, the probability of the repayment of
principal and interest of an obligation is measured by means of credit rating. A higher credit
rating shows a low credit risk. The assessment is realized by a rating agency. According to the
assessed object, credit ratings of the state, company, municipality, financial institution, single
bond, etc. exist. Credit rating is a result of a credit rating process. It is represented by the j-th
rating class ωi,j∈Ω, Ω={ω1,j,ω2,j, … ,ωi,j, … ,ωn,j}, where n stands for the number of objects
and Ω is a rating scale. The rating class ωi,j∈Ω is assigned to the i-th assessed object oi∈O,
O={o1,o2, … ,oi, … ,on}. Based on the above, the credit rating modelling is considered to be a
classification problem with the aim of classifing the i-th object oi∈O into the j-th rating class
ωi,j∈Ω.
Credit ratings are used by bond investors, debt issuers, and governmental officers as a
measure of the risk of a company. They provide a means of determining risk premiums and
marketability of bonds, allowing firms issuing debt to estimate the likely return investors
require. Bankers and companies considering providing credit rely on credit ratings to make
important investment decisions, many regulatory requirements for financial decisions are
based on credit ratings, and some companies are restricted to investment grade bonds.
Credit ratings are costly to obtain because rating agencies invest large amount of time and
human resources to perform the credit rating process. Therefore, there has a large much effort
made in order to simulate the credit rating process of rating agencies through statistical (e.g.),

 and soft-computing methods (e.g.). The difficulty in designing such
models lies in the subjectivity of the credit rating process. This subjectivity is emphasized as
the mode in which complex relations between financial and other variables are evaluated.
Such a complex process makes it difficult to classify rating classes through statistical
methods. However, soft-computing methods (neural networks, fuzzy systems,
evolutionary algorithms, artificial immune systems, and hybrid systems) can be
applied for the modelling of such complex relations.
Therefore, soft-computing methods, so far, have been used for corporate credit rating
modelling. As a result, high classification accuracy has been achieved by neural
networks and support vector machines (SVMs). In economics and
finance, neural networks (including SVMs) are usually applied in such cases where variables
are in non-linear relations. This is reported to be typical for economic and financial data.
Neural networks make it possible to model these relations as they learn the dependencies in
training data. As a result, gained knowledge is stored in synapse weights. Moreover, the
knowledge can also be applied for unknown input data which were not used in the training
process. This is also known as generalization ability of neural networks. The disadvantage of
neural networks lies in the fact that neural networks are usually designed as so-called “black
boxes”, i.e. it is difficult to extract understandable knowledge from them. Therefore, prior
studies in modelling credit rating are aimed at quantifying the effect of input variables for
classification, i.e. to find out which input variables are crucial for credit rating process.
Mostly, sensitivity analysis has been employed for this purpose. Based on the mentioned

Finance and Banking Developments (Banking and Banking Developments) by Charles V. Karsone