Industrial applications of artificial intelligence a survey from the past to the future Essay

The banking transactions passed through several stages with different attempts to speed up the processes. In a study for (Altman, 1968) this study explained several traditional measures by using discrimination analysis for credit scoring, The banks used to assess the credit worthiness by qualitative information, mathematical and statistical techniques, The traditional approach applied in 1930 was to evaluate only a qualitative measures only information about the corporation that apply for a loan without considering any financial challenges for that institution or the probability of insolvency, Banks then realized that it is not logical to apply behavioral science for assessing credit the second discrimination measure is called multiple discriminate analysis MDA is a statistical method used to discriminate by following the characteristics of the loan applicant by knowing which group this person belong to. This approach is constructed by identifying several groups and evaluation is done based on the group to which the applicant belongs to, for instance statistics identify that male group in a specific area have a high probability of default so if the loan applicant is man and from the mentioned area so it is a signal for bank that this person may default, This method has been applied to classify high and low income level for both individuals and firms.

In 1977 a new discrimination model was done by (Altman, et al., 1977) called ZETA this model is to identify the bankruptcy of a corporations by using statistical measure called multivariate. According to (Altman & Saunders, 1998) the use of this model is to minimize to the variance among the same group. All previously mentioned qualitative and techniques was the traditional and first intuitive towards improving the efficiency of banks in loan granting and credit scoring these tasks were done by human without machine intervention.

In 1997 an invention of a software to automate the process of granting the loan from the first step till final approval according to the inventors (Dykstra & Wade, 1997) in traditional methods the applicant will not only face the problem of being discriminated based on the group to which the person belongs to but also the applicant has to do many manual transactions and go through long processes starting from filling the applications go to the Certified public accountant CPA, paying a lot of fees and finally waiting to the reply from the bank side where also a lot of steps and human involvement until issuing the final decision of granting the loan or not. How this system works? The answer for this question it is a kind of user-friendly system as each user has a credentials can be used for access for the lending institution, then the applicant can fill all the required documents, it is important to mention that each institution has an acceptable score in other words if the score has been calculated below the required institution needed score the loan will be automatically rejected.

Based on the previous literature reviews this paper summarized the previous studies in the following table. This table is a summary for the development of techniques to grant the loan.

Table 1.1

Date The technique of assessing the loan worthiness.

1930 The decision for granting the loan based of qualitative measures.

1977 Application of statistical approaches such as multiple discriminate analysis MDA and ZETA.

1997 First software invention to automate the process

The last loan processing approach was the first approach to consider Artificial intelligence in credit scoring and assessing the credit worthiness of the applicant. It is argued by (Chatterjee, et al., 2000) AI made a massive growth in areas concerning mimic human actions in the past it the uses of AI was very limited, the modern approach of AI is imitating human brain and producing output based on massive knowledge provided to the computer this is called neural computing or Artificial neural networks ANN. but before going into details of this technique a definition of neural networks should be mentioned. According to (Malhotra & Malhotra, 2003) neural networks is a simulation for human brain nervous system as human brain receive an input and then start the processing process to get an output. the same thing will be applied for banks but by using different input of income, applicant previous payment and many of the applicant characteristics to get an output of either loan acceptance or rejection recent studies used NN for credit and debit card fraud detection.

It is stated by (Bahrammirzaee, 2010) that Artificial neural networks not only for assessing credit worthiness but also the application of ANN could be in different fields such as assessing behavior attitudes and evaluating the financial statues.

In a study for assessing the performance and efficiency of banks that apply AI by (Fethi & Pasiouras, 2010) this study includes previous 196 studies published from 1998 till 2009 the study concluded that banks that apply AI perform better than banks that still apply traditional techniques in credit scoring, the study also mentioned the application of ANN perform better in expecting bankruptcy than credit scoring.

In a study for showing different uses of Networks by (Maes, et al., 2002) the study proves the efficiency of networks in other fields of the banks such as credit and debit card Fraud detection but there are other techniques of using networks differs from ANN is called Bayesian Belief Networks BBN, One of the problems that bank may face with customers is charging them with high fees or high interest rate, but in many cases the customer claim that he should not pay that extra fees because of fraud happened in other words somebody was able to get his pin code in this case it is the responsibility of banks to start the process of fraud detection. This study concluded that BBN performs better than ANN in this field of detecting fraud.

2.2 Application of AI in Anti-Money Laundering

In a conference for money laundering and uses of artificial intelligence for anti-money laundering (Gao, et al., 2006) defined money laundering as the process of changing the money sources appearance that gained from trading drugs and tax avoidance from dirty to clean in other words from illegal to legal. The amount of money that has been gained from this process represents 2 to 5 percent from annual Gross Domestic process. According to this paper the banks used to assign a certain value or threshold they start to follow the transactions of this account if the money exceeded that assigned value, but this method is easy to avoid by identifying that value and process with transactions with values lower than threshold for examples , by investigating the sources of finance for 09-11-2001 events small amount of money has been transferred which is below the threshold so the bank was not able to detect that transactions. According to (Gao, et al., 2006) the application of AI in this field will be through intelligent anti-money laundering system (IAMLS) by following each transaction that is irregular ",customer behavior ",and adding customer usual inquires to be able to group this account either a suspicious or normal account, the design of this automated system is to collect data ",monitoring the agent and diagnose his /her behavior by adding all that data and features to the system all mentioned data will be processed by the system to generate a report for this client. The next section will analyze how that modern approach will reshape the structure of banks.

2.3 Reshaping the future structure of banks by AI

By adding more technology this represents a threat for employees this is the definition of technical change as stated by (Acemoglu, 2002). In this study it is worth to mention the effect of what previously mentioned on working force and what employees in banking sector will face by applying that advanced techniques.


according to previously mentioned statistics by (Crosman, 2018) there is a threat of using AI in banking sector around 1.2 million people will lose their jobs by 2030 most of them are tellers as the automated machines can easily replace their daily tasks.

How to cite this essay: