Analysis of Credit Operations Using the Bayesian Network - Paper Example

There is no industry that has no risks. However, risks can be minimized, managed, shared, accepted, transferred but cannot be ignored. (Lathma, et al. 1994). The banking sector is regarded as one of the most uncertain industries as the credit operations is one of the riskiest kind of work (Iturriaga, et al..2015). The complexity of the banking industry and its credit issuing procedures generates greater risks. The risk is associated with the credit operations against the collateral. These risks can mostly be incurred at the early stages of credit granting though can even happen in later stages (Tuckwell, et al..2013). Credit operations against collateral are believed to be one of the greatest barriers to organizations meeting their mission due to the changes they cause. Risk management and assessment is an important component of the banking industry and other industries as well. It is a continuous process of risk analysis, risk identification, risk review and monitoring as well as risk treatment (Aglietta, et al..2014). Among these, the risk analysis is the most challenging as it will entail the risk assessment. Despite being the most challenging, it is regarded as the most important part of the risk assessment (Wolff, et al..2014). This thesis will put a lot of focus on the methodologies that can be used as decision support system for risk assessment for credit operations against collateral. Various researchers have looked into several techniques and theories for assessing the credit operations risks. They have brought fought some decision support system and range of software, but the proposed tools have failed because of the low take-up in practice.

The banking industry has a lot of risks arising from credit operations. Credit operations against collateral is a major risk that any bank must have in place strategies or technologies that will help solve the situation (Kumar,et al.2015). It is good therefore to have a decision support system for risk assessment of the credit operations against collateral (Abdelmoula,et al.2015). This will help to minimize any kind of loss that can be suffered as a result of credit that has been issued by any organization. In most of the organization system, in credit decision making, there is the probability of default models that finds out the cost of capital and in price agreement (Kelliher,et al.2017). Furthermore, most of the central banks have dramatically evolved to a setting where the use of these models works to attain firmness standards for credit risk valuation in the bank system (Srinivas, et al.2015). In the banking industry, the credit risk assessment normally depends on the credit scoring models. The guarantors during the credit are divided into two main parts (Hautsch,et al.2014). These are the guarantor with the assets and the guarantor with the individual. All of this focus to enhance security in the credit operations as a payment alternative to the credit holder that is given to the lender ad cannot honor to pay the credit on the agreed time (Sousa,et al.2015). On the part of the creditor, it gives the liquidity security from receiving operations. The admeasurement of the recoverability of credit is systematic that looks into the efficiency of the mechanism of return of capital invested in collateral (Hossain,et al.2015). The guarantor of the assets gives a high operating cost, and it is challenging to follow in loco for the evaluation procedures an asset. This explains why there should be a specific attention owing to the large volume of goods and insufficient technical and operation capacity. (Sadatrasou, et al.2015). This thesis will look into details various methodologies of the Decision Support System for Risk assessment in credit operations against collateral.

Analysis of credit operations using the Bayesian Network

Credit operations

Credit operations are most frequent in the banking industry. It is therefore important to analyze the DSS for assessing the risk factors on credit operations against collateral in the banking industry. Bank credit risk analysis is largely used in all banks globally. Therefore, the credit risk analysis is very critical and also the dreary process (Hossain, et al.2014). There should be a variety of risk methods that should be used for assessing risk. Also, credit risk is one of the main functions of the banking industry Organizations such as banks categories the customers based on their profile. It may range from the financial background of the customers to the subjective factors of the customers (Sadatrasou,et a.2015). Financial ratios play a very great role for the risk level calculation. The financial ratios can show the financial statement of the firm as they are objective. Income statement and cash flow are some of the financial statement that can be used to calculate financial objective financial ratios. Subjective factors can also be used to calculate the risk level factors depending on the mission of the bank and the bank decision strategy (Lalon,et al .2015). Credit scoring model can be used to show both subjective and objective factors. The quality of services surrounding the credit operations is very crucial. This is because of the organizations income as well as the market share of the firm, lets says the bank, and is directly related to the granted credit. This section has a summary of various studies for the risk analysis with the Bayesian Network decision tool. It will also show how it can be used to design a decision support system from the cause-effect relation as well as the conditional probability.

2.2 Bayesian Network, there Neural Networks and Logic regression

Developing an idea of decision support system in the credit scoring domain through the Bayesian Network can be very sufficient (Mei,et al.2016). Applicability of Bayesian network in the areas of the process of the working capital credit scoring is very essential in the risk assessment. The Neural networks can be used to make credit risk evaluations. This is because they put a lot of emphases the significance of universal approximation property and the high prediction accuracy that is associated with this model (Wu,et al.2015).. This decision tool when also used will be able to show the negative ways that are in the evaluation. However, it is challenging o understand how the conclusion have been reached at (Wahyudin,et al.2015). This can best be explained through the credit-risk data and analysis done by the WEKA software and the NeuroRule extraction technique. During the design of this decision tool, the Netica software package is used to create the Bayesian network (Triki,et al.2016). This means that both Netica software and the WEKA software can be used.

Through using the data mining, it can be easy to develop a predictive defaulter mode. Developing credit-risk evaluation expert system using the neural network rule extraction and the decision tables can be very effective (Bayrakdaroglu,et al.2016). This is because the neural networks do very well regarding their performance for the complex and unstructured problem when they are compared to the traditional statistical approaches. This makes them able to attain a high predictive accuracy rate and the reasoning behind how they come about their decisions is not readily available (Firoozye,et al.2016). Some of the techniques that can be used in this decision tool when used in a DSS are the logistic regression, Discriminant Analysis, Classification and regression tree as well as the neural networks.

2.3 Risks probability when evaluating results to determine the risk assessment of the credit operation

Decision theory majorly concentrates on identifying the uncertainties, values as well as other essential issues that are relevant to a given decision relating to the issuing of credit as well as the resulting optimal decision. The decision theory gives the basis for thinking about the challenges of an action (Chorniy,et al.2015). Therefore, while making the decision, there is a great consideration in the probabilistic values for the random variables or each event (Witzany, et al.2017). The random variables or each event are the subjective factors, financial ratios, credit risk classes or the financial, economic classes (Srinivas, et al.2015). They can either be conditional probability or prior probability.

Conditional probability: When extra information is provided when the prior information is already known, the conditional probability can be calculated. The extra information can be collected through observation (Tesfaye,et al.2013). For instance, in the calculation of the financial ratios, there is additional information that the banker gets. This probability can be represented as P (x/ w) (Goodfriend, et al.2014) Decision rule is similar to the prior probability (Hasan,et al.2017). The greatest conditional probability is chosen. If some financial ratios are calculated, the descendant nodes probability tables are changed with regards to the known conditions (Wahyudin,et al.2016). For instance, in assessing the risk factors that are associated with the credit operation in a bank, the bankers are required to determine the credit risk of the client. The probability table used in banks as per the study conducted by Baesens, 2016 for the credit risk class is shown in the table below.

Credit Risk Class

One 53, 7%

Two 25/3%

Three 12/5%

Four 7/4%

(Baesens,et al.2016)

From the table shown above, the classs one, two, three and four stands for the risk classes with class four being the worst and class one being the best. Normally, the first and the second classes are always approved. However, this will depend on the strategy of the bank (Ansah,et al.2013). There are some institutions that can improve all the classes with the intent of boosting their market share (Ntwiga,et al.2016). The risk that an institution will subject itself to will depend on the number of the lower class credit approvals that they have approved (Henry,et al.2013). When they have approved a lot of such credit from the lower class, they are likely to have an increase in the credit risks. According to the decision theory, the highest conditional probability is selected. Therefore, the credit class one is chosen.

Prior probability: In this kind of probability, it is assumed that there are some prior probabilities. The probabilities are calculated from the previous data (Margulies,et a.2016). For instance, when one wants to calculate the prior probability of the Average Collection Period that are the financial ratios to indicate the roughly amount of the period that it takes for an organization to receive payment, the formula that is used is: Average Collection Period = (Days * AR)/Credit sales (Pai,et al.2015)

The values are available in the financial statements of the firms. The ratio is calculated for every customer (Weber,et al.2014). The results for all the customers are then classified depending on the strategy of the firm or the bank. In most circumstances, the financial ratios are categorized as good, medium, bad and very bad (Schuermann,et al.2014). Each bank or organization has their range under which they will classify a particular value.

2.3 Analysis of the Bayesian network

The Bayesian Network which is also known as the Belief Network is a graphical decision mode...