Expert Systems

Expert Systems

Expert Systems are computer systems which embody some of the experience and specialised knowledge of an expert and thereby mimic the expert and act as a consultant in a particular area. Knowledge is often represented in an Expert System in a knowledge base, which is a network of interconnected rules which represent the human expertise.

Expert Systems fall into a field known as Artificial Intelligence. This field is concerned with the development of computer intelligence. The goal of Artificial Intelligence is to develop computers that can think, as well as see, hear, walk, talk and feel. A major thrust of Artificial Intelligence is the development of computer functions normally associated with human intelligence, such as reasoning, learning and problem solving.

Advances in computer processing power open the way for wider use of so-called Artificial Intelligence, at the same time that the self-serve aspect of online processes has increased the need for systems that "think." As time goes on and as patterns change, the computer learns because it is constantly being given cause and effect.

The Expert System

	User

			Knowledge Engineering				Expert

Expert System Development

Financial Decision Making

The financial services industry has become a vigorous user of Expert System techniques. With their high level of accuracy and reliability, Expert Systems easily replace complex human decision-making. This is done with the help of their expert knowledge-bases into which data and logical reasoning are inputted. Expert Systems technology is normally introduced to bring consistency to the decision-making process, as well as to speed it up through semi-automation.

Expert systems are increasingly being used in banking applications. Some of the areas where this technology is being applied are credit risk analysis, portfolio management and analysis, credit approval, loan application processing, money transfer processing, loan monitoring, teller training, financial planning services, and intelligent message understanding. These applications usually have a clear set of rules and regulations to be followed which can be implemented as an expert system. In other areas, for example foreign exchange trading, securities trading, strategic planning, securities analysis, and specialized financial planning, there is expertise that is truly unique and not available in a written format as a well-defined set of regulations.

Advisory programs have been created to assist bankers in determining whether to make loans to businesses and individuals or whether to reject their applications. Various decisions have to be taken in such a situation before arriving to a definite answer that should prove to be the most profitable to the business and could save the same organisation from undesirable bad debts.

Insurance companies have come to use Expert Systems to assess the risk presented by the customer and to determine a price for the insurance. Insurance Fraud Prevention techniques are also widely utilised and give an important helping hand to management and the company as a whole in their fight against fraudulent customers.

Money Laundering Prevention could also take a giant's step forward with the use of an appropriate Expert System. Cash transactions would be controlled and each banknote checked independently to assess whether it is a valid banknote or a forgery. This could, and should, discourage prospective developers of false money from circulating their counterfeits. Neural networks make it much more difficult for fraudulent activity to go undetected.

A very interesting issue, as already mentioned, is the one concerning bank managers and their crucial decisions on what amounts to lend, if at all, to customers.

Improved objective decision-making

Extensive analysis has to be carried out with experts in this delicate field to develop the business rules before building the Expert System.   As there is a significant element of subjectivity in this kind of decision-making, the most difficult part of developing the business rules was determining what factors to take into consideration when making a decision. The final aim of the system would be to provide a recommendation, based on the bank's standard lending policy, on whether to loan to the client or not, and the size of the amount.

The most obvious point to start is the reason provided by the customer asking the loan. A comprehensive list of reasons used was compiled from archived applications and each is given a weighting. Tolerance levels could be set to categorise the reasons as "disaster", "exceptional", "normal" or "trivial" depending on the total weighting score.

Still, up to this point, nothing can be decided because being in a difficult financial situation and needing money badly does not mean one is entitled to receive a loan by the bank. A serious banking organisation will look deeper into the matter and fish out more details about its prospective customer.

The borrowing history of the customer in consideration could be a good place where the bank could concentrate. The person's lodgement patterns are assessed as being "good" or "bad". The taxpayer's lodgement history is given a score dependent on whether previous returns have been paid on time.

Using a neural network, a bank feeds the computer data, such as the characteristics of loan applicants' income, how many kids they have, and where they live. The computer is also given a description of the type of customers who defaulted on loans in the past. A neural network can process quantities of data to discern a relevant pattern beyond the capabilities of other systems. Other factors include a borrower's debt-to-income ratio, employment status, the length and type of loan, the reason why a loan is being asked for, other parties involved in the loan, past bankruptcies, and any mortgage delinquencies.

So for borrowers with good credit, the automated system allows higher debt-to-income ratios than other customers with less of a reputation. That means a borrower might qualify for a larger loan than someone with the same income and poorer credit.

For example, a reason such as "Purchase of a private residence" would be given different ratings according to the customer offering that reason. If the customers are a couple of newly-weds looking for their first home, the bank might check their parents' previous borrowing records, if any. The occupations held by the couple might also prove relevant to get an idea of after how long the loan might be repaid. What type of residence the couple wants is also significant. If they desire to purchase a small flat the amount they require will be much smaller (and much more easily repaid) than if they are planning to purchase a bungalow.

The Expert System will reserve a different treatment to a twenty-year old customer asking for a loan to build his own discotheque in his attempt to attract youngsters to a new leisure area. Before committing itself to lending this youngster money, the Expert System must be sure that the client will be able to repay. Due to the young age of the prospective client, the checks carried out might be more thorough. The young age might put the youngster in a bad light but if he has financial backing from his parents he might be able to receive the loan.

As already stated, the lending sector is a very delicate one and needs a lot of subjective judgements. Artificial Intelligence of Expert Systems may not be enough to fully automate such a system and therefore human intervention is a must. When making decisions on whether or not to dispatch a loan to a client, such details as the bank's reputation are to be taken into account. A too strict lending policy may result in a deterioration of the bank's popularity with clients and hence result in loss of business. Such aspects cannot be measured objectively, and in such an aspect, human resources should amalgamate their knowledge with the one provided by the Expert System so as to come out with the best possible solution both for the bank itself and also for the client demanding the loan.

This does not mean that the Expert System is an infallible way to secure that the loans given out are repaid in due time and without any problems. Even though Artificial Intelligence has improved immensely these last years, it is still an artificial way of arriving to a particular result. Human intervention is still an important resource which even an Expert System cannot replace completely. An expert and professional manager can assess situations in a different manner since he can actually talk to the clients and can draw conclusions which an Expert System would not be able to determine. This fact is more strengthened by the fact that this particular banking sector involves millions of pounds and cannot be managed in an irresponsible manner. Although the margin of error of Artificial Intelligence may be minimal, this should further be reduced by involving human intervention to examine and counter-check the work done by Expert Systems. After all, interaction between client and the bank must exist in one way or another. Delicate matters such as money borrowing need to be based on a reciprocal trust between both parties (in this case the client and the bank representative) and Expert Systems, however accurate, cannot provide such interaction.
Decision Support Systems

A Decision Support System is a type of Information System whose principal objective is to support human decision-making in the circumstances that human judgement and computer processing are needed. Decision Support Systems are generic in nature therefore they have a wide coverage. They are extremely useful because by simply altering their variables, they can emulate diverse scenarios. They assist managers with unique, non-recurring strategic decisions that are relatively unstructured. DSS, therefore, tend to support tactical and strategic decision-making in situations where the risk associated with any error is high and a mistake can have serious consequences. It is therefore important to consider all alternatives and to evaluate them carefully.

A Decision Support System too could have its various advantages in its utilisation for credit approval in banking, but most probably it would not be as powerful and as influential as an Expert System.

A Decision Support System could provide correct and supportive results to the manager, but would not be able to learn from past experience so as to provide an even more accurate result when called into action again. Contrarily, Expert Systems can learn from experience, and increase their knowledge. Thus, the next time around the Expert System could provide even more accurate results.

Conclusion

Unlike Decision Support Systems, Expert Systems provide answers to questions in a very specific problem area by making human-like inferences about knowledge contained in a specialized knowledge base. They must also be able to explain their reasoning process and conclusions to a user.

The main objective of a Decision Support System is assistance to human decision making, while for an Expert System, emphasis is based on replication of the reasoning of a human advisor. While in a Decision Support System, it is the user who queries the system, in an Expert System, the roles are inverted. It is the system who queries the user and takes decisions according to the inputted data.

The major limitations of Expert Systems arise from their limited focus, maintenance problems, and development costs. Expert Systems excel only in solving specific types of problems in a limited domain of knowledge. They fail miserably in solving problems requiring a broad knowledge base and subjective problem solving.

However, the advantages brought about by these powerful packages by far outnumber the disadvantages. Expert Systems and the vital field of Artificial Intelligence are going from strength to strength. Progress continues and only time will tell if the ambitious goals of Artificial Intelligence will be achieved and equal the popular images found in science fiction.