EXPERT SYSTEMS

Expert systems have evolved out of the work on artificial intelligence over the past few
decades and are finding increasing applications in business. The system gathers together a
database of knowledge or expertise to offer advice or solution for problems in a particular area
by emulating the abilities and judgments of human experts. It accumulates all the expert
knowledge in a given area so that the advice or solution offered is better than that of a single
consultant or expert. It guides users through problems by asking them a set of questions about
the problem. The answers given are checked against the rule base in the system to draw
appropriate conclusions from the problem situation. Expert systems are particularly useful in
dealing with unstructured problems.

Expert system was originally developed to replicate abilities of human experts. The
system captures and stores human knowledge in a area of expertise, called domain, a uses it to
solve problems which other wise requires the help of human experts. The solution suggested by
the system is expected to be superior to that by any single expert.
Expert systems are designed to solve real problems in a particular domain that normally
would require a human expert. It can solve many types of problems. It is designed to solve
some problems very effectively. But it cannot solve every problem one might encounter in an
area.
Developing an expert system involves extracting relevant knowledge from human experts
in the area of problem, called domain experts. Such knowledge is often heuristic in nature. That
is, it is some useful knowledge based on some “rules of thump” rather than absolute certainties.
Acquisition of such rules of thump and storing them in knowledge base are serious tasks in
building a knowledge base. A knowledge engineer does this of knowledge acquisition and
building a knowledge base.

The expert system consists of two major parts: the development environment and the
consultation environment. The expert system builder uses the development environment to build
the components and store expertise into the knowledge base. Non-expert user uses the
consultant environment to get the expert opinion and advice from the expert system.

COMPONENTS OF EXPERT SYSTEM

 The expert systems have the following components:
(1). Knowledge acquisition facility
(2). Knowledge base
(3). Knowledge-based management system
(4). Reasoning capability
(5). Work space
(6). Explanation facility
(7). Inference engine, and
(8). User interface
These components are briefly explained below:
1. Knowledge Acquisition Facility
Domain experts acquire expertise in their area of expertise over a long period. The
expertise may be the result of their constant interaction with similar experts, observation and
personal experience in the domain. Capturing expertise is one of the most difficult tasks of
building knowledge base. This facility adds new knowledge and rules to the existing knowledge
base and ensures its growth to meet emerging need. Usually a knowledge engineer takes care of
this task. He identifies and interacts with the domain experts to gather expertise.
2. Knowledge Base
Knowledge base is just like the database of information system. It stores knowledge and
rules and explanations associated with the knowledge. Knowledge representation is a major task
in expert system building. The knowledge must be meaningfully represented in the system so
that the system can relate to real world problems.
The knowledge base includes three of knowledge such as:
Factual knowledge
Heuristic knowledge, and
Meta knowledge
The factual knowledge consists of facts about the domain, say, finance, medicine, design,
etc., heuristic knowledge relates to the rules associated with a domain or problem area. Meta
knowledge enables the expert system to use and analyses facts, extract those facts and specify the
route to a solution. It refers to the ability of an expert system to learn from its own experience.
The knowledge base contains data and facts relevant to a problem area. The most
common way to represent knowledge in expert systems is in the form of rules such as if
……then statements. Semantic networks and frames are other forms of knowledge
representation in expert systems. The inference engine contains reasoning methods. It is a piece
of software that probes the user and searches the knowledge base for the appropriate solution.
The user interface links the user with the expert system. It sorts up screens for user-interaction
with the system. Such interaction leads to identification and solution of problems.
3. Knowledge-based Management System
It is similar to a database management system in an information system. Its major task is
to up data the knowledge base with knowledge and rules.
4. Work Space
The workspace or black board is a memory area used for describing the current problem,
and storing intermediate results.
5. Explanation Facility
Most expert systems have explanation facilities. It explains how recommendations are
derived. The user can know how the expert system arrived at the solution, why some alternatives
were rejected, why some information was asked for etc. The explanation facility answers these
questions by referring to the system goals, data input and the decision rules. For example, in
case of loan proposal evaluation, the expert system’s explanation facility will clarify on probing
why one application was approved and why another was rejected. In case of a medical expert
system such as Mycin, this facility builds confidence in the user about the expert system and the
solution it provides to problem.
6.Reasoning Capability or Knowledge Refinement
The expert system has the capability to analyze why its solution failed or succeeded and
ways of improving its solution.
7. Inference Engine
The inference engine works like the model base in decision support system. It
manipulates a series of rules using forward chining and backward chaining techniques. In
forward chaining the inference engine poses a series of if ….then condition checking. Based on
the responses a particular solution is suggested. In backward chaining technique, the inference
engine starts with the goal and checks whether the conditions leading to that goal are present.
8. User Interface
The system provides an interface for the users to interact with the system to generate solutions.
It is similar to the dialogue facility in decision support system. The artificial intelligence
technology tries to provide a natural language interface to users.

COMPONENTS OF EXPERT SYSTEM – AN ALTERNATE APPROACH

Another way to analyze expert system components is to look at the physical ingredients
of the system as follows:
1. Hardware
Expert system shells operate on all types of hardware such as micro-computers,
minicomputers and mainframe computers. Since microcomputers have become ubiquitous, it
has become the standard hardware platform.
2. Software
Two types of software are needed for expert system: symbolic programs and expert
system shells.
The expert systems were developed in symbolic programming languages. They have
facility to process symbols rather than numbers. Expert system shells consist of a set of
programs that provide an environment to declare rules and other knowledge. The shell also acts
as an interface between the user and the expert system.
3. Knowledge
Expert system stores knowledge for decision-making. The knowledge may be
represented in a meaningful way such as rule-based format or frame-based or a combination of
both so that it can be retrieved to solve real-life problems.
4. People
Expert system requires some personnel with some expertise.
5. Knowledge Engineers
Knowledge engineers are responsible for creating an expert system. They interview
domain experts and build knowledge base of the system. Their tasks are acquiring knowledge,
modeling knowledge, and encoding knowledge.
6. Procedures
Both the users and expert system operators have to follow certain procedures for working
with the expert system. The procedures for normal operations and recovery operations have to
be developed and maintained.

CHARACTERISTICS OF EXPERT SYSTEM

The following are some of the characteristics of Expert system:

• Expert system is capable of handling challenging decision problems and delivering solutions. Expert system uses knowledge rather than data for solution. Much of the knowledge is heuristic-based rather than algorithmic.
• The knowledge is encoded and maintained separately from the control program.
• Expert system has the capability to explain how the decision was made. It can also state why a particular piece of information was needed for the solution.

WHEN AN EXPERT SYSTEM CAN BE USED?

An Expert system can be used if the following conditions are met:

• The problem cannot be specified in terms of a well-defined algorithm.
• The problem requires consistency and standardization.
• The domain or problem area is narrow or limited.
• When the task is hazardous.
• There is scarcity of experts in the area.
• The problem involves complex logic or a large number of rules.
• Human experts have successfully solved similar problems.

ADVANTAGES OF EXPERT SYSTEM

Expert System is an application area of artificial intelligence. Its purpose is to analyze how
human experts make decisions and replicate this decision capability at affordable costs for
organizations. It advantages are:

• It enhances decision quality.
• It reduces the cost of consulting experts for problem solving.
• It provides quick and efficient solutions to problems in narrow area of specialization.
• It offers high reliability of expert suggestions or decisions.
• It gathers scarce expertise and uses if efficiently.
• It can tackle very complex problems that are difficult for human experts to solve.
• It can work on standard computer hardware.
• It can not only give solutions, but also the decision logic and how the solution was arrived at.

Hence, the explanation facility permits a review of the decision and its logic.

LIMITATIONS OF EXPERT SYSTEMS

Expert System is not the result of a one-shot development. It is subjected to an iterative
process of problem identification and refinement. Once the system is found to be working
satisfactorily, it is implemented for consultation and decision-making. Some of the limitations of
expert system are:

1. It is difficult to extract expertise from human experts and hence the knowledge base may not be complete. 
2. Each problem situation is different from problem to problem. Hence, the solution suggested by a human expert is bound to be different from expert system solution.
3. Expert system is effective in solving specific problems in narrow domains. It fails in properly analyzing problems in a larger area and in suggesting solutions.
4. The cost and time required for developing expert system are very high. Hence, expert system is not affordable for most firms.
5. Expert systems are expensive to build and maintain. In many cases the system has to be developed for the organization. If packages are installed they may have to be customized for the requirement of the organization.
6. It is impossible to build any useful expert system as the expert system cannot capture all the assumptions on which real-life decisions are based.
7. It takes long period of time to develop and fine tune an expert system. 
8. Large expert systems are difficult to develop and maintain.

EXAMPLES OF EXPERT SYSTEM

Business Insight
This expert system offers facility for strategic analysis. It is based on knowledge from
over thirty business experts. It identifies strength, weaknesses, inconsistencies etc. and gives a
through explanation. Business Resource Inc. developed this system.
Forecast Pro
It is a business forecasting expert system. The expert system examines data and the
program helps the users in forecasting using statistical tools like exponential smoothing,
regression, Box-Jenkins etc. This software is from Business Forecast System.
Prospector
It is designed for use in prospecting for minerals. It recommends probable sites for
mineral deposits.
Mycin
It is a famous expert system for medical diagnosis. It diagnoses blood infections and
recommends treatment.
Expert system is an extension of the artificial intelligence. The system is designed for
mimicking experts in there are of expertise. It collects and stores knowledge and rules in its
knowledge base. A user interface is provided to the user to interact with the inference engine in
generating solutions and getting explanations. Expert systems are useful in tackling complex
problems without the help of experts. They are used in a wide variety of applications such as
diagnosis, design, planning, forecasting and control. Even though it is extremely useful for
decision-making, it has its own limitations.

DECISION SUPPORT SYSTEMS

 A decision support system is a computer application that helps users analyze problems
and make business decisions more confidently. It uses data routinely collected in organizations
and special analysis tools to provide information support to complex decisions. For example, a
firm’s sales department may be interested in analyzing various sales decision options. The
decision support application might gather data, present the data graphically and help in
evaluating various options. It may use past sales figures, project sales based on sales
assumptions for each alternative considered and display information graphically. It may also use
artificial intelligence to enhance its decision support capability.

Decision Support System assists managers in making unstructured decisions. The system
enables them to interact with the database, model base and other software. It enables the users to
generate the information they need rather than depend on some reports produced according to
some anticipated information needs. DSS is more suited to handling unique and non-routine
decision problems. In many situations the problem itself may not be easily identified. Similarly,
identifying alternatives, identifying outcomes of each alternative considered, evaluation of
alternatives etc. pose problems to the decision maker. Each problem might require a different
approach to problem definition, analysis and resolution. Not only that it is difficult to solve such
problems, it is also possible that the decision process and solution vary with the decision maker.

DSS is designed to support managerial decision-making, usually, at middle and top levels
of management. Decisions made at the top level are mostly futuristic and non-repetitive in
nature. Such decision situations are highly uncertain and even specification of information
requirements for decisions are difficult. They are classified as non-programmable or
unstructured decision situations. The impact of such decisions will be seen throughout the
organization and cost of a wrong decision is usually very high, for example a decision to sell off
a line of business. This is in sharp contrast to programmable or structured decisions where the
decision procedure can be well defined and every information requirement can be pre-specified.
Most of the decisions taken at lower levels of management fall into this category. For example,
a decision to replenish stock of an inventory item is a highly structured decision taken at the
operational level. DSS is intended to help managers making unstructured decisions. ‘the system
includes a database, various models (Mathematical models for optimization etc.) and an interface
for the manager (usually a terminal) to interact with the system. The manager takes data from
the database, selects appropriate model or models and analyses the data using these models to
know the probable results of various actions.
DSS is thus an interactive computer system with many user-friendly features aimed at
helping non-computer specialist managers in making plans and decisions on their own. With the
recent advances in computing technology, particularly the powerful microcomputer and
interactive devices, the uses of DSS is expanding rapidly as these managers find it easy to access
databases and model base for retrieving and analyzing data.
DSS contains a database, models and data manipulation tools to help decision makers. It
is useful where decisions are semi-structured or unstructured. The decision rule for a structured
decision can be pre-specified. Hence, it is possible to automate such problem solving.

Intelligence activities are targeted at discovering problems of organizations. The
information reporting system can handle most of these information requirements. In the design
phase, alternative solutions to the problem identified are generated. This stage requires more
focused information and more intelligence based systems like DSS and Knowledge based
systems. Choice phase involves selecting the right alternative. This requires thorough
evaluation of the consequences of all the alternatives under consideration in terms of risk and
return, and its impact on problem area.

Decision implementation is a critical phase. Managers are anxious about the results of
decision implementation right from day one of implementation. In this phase, managers call for
information on implementation of decision such as stage of implementation, time and cost
involved, implementation constraints, and impact of implementation.
DSS can support repetitive or non-repetitive decision-making. It provides capabilities
for repetitive decision-making by defining procedures and formats. For example, an insurance
agent may use a DSS package to help clients in choosing insurance schemes. With the
privatization of insurance in the country, innovative insurance products are being introduced. An
investor will find it difficult to properly identify an insurance product matching his or her
requirements. The agent can carry a laptop with a DSS for insurance products to his clients. The
DSS can be used by the sales agent to demonstrate to the clients the details of each scheme in
terms of risk covered, bonus, maturity value, premiums etc. and help the clients arrive at their
decisions to purchase insurance policy.
DSS can also help non-routine decision-making. In fact its utility is high when nonrepetitive
decisions are made. For solving a non-repetitive problem, the DSS provides data,
models and interface methods to the user to select and analyze data. For example, a marketing
manager might want to analyze the potential demand for new products that the company is
planning to introduce. The marketing manager can use a DSS to forecast the demand using
relevant data about the market obtained form some database service firms like Centre for
Monitoring Indian Economy. The analysis will provide new insights into the market behaviour
and product performance that will help the manager in introducing new products into the market.