Uncertainity in Expert Systems and Various

Techniques used in Handling them

Tanmay Baranwal

Ravi Kiran Boggavarapu

Himanshu Singh

Section/Roll : KE-031/R08
Course Code : CSE508M
Lovely Professional University
Phagwara, Punjab

Section/Roll : KE-031/R01
Course Code : CSE508M
Lovely Professional University
Phagwara, Punjab

Section/Roll : KE-031/R07
Course Code : CSE508M
Lovely Professional University
Phagwara, Punjab

AbstractThe aim of artificial intelligence is to develop tools

for representing piece of knowledge and providing inference
mechanism for elaborating conclusion of knowledge from stored
information. The available knowledge is far from being certain,
precise and complete. In Expert systems the word uncertainty is
related to the working with inexact data, imprecise information,
handling identical situation, reliability of the results etc. An
expert system allows the user to assign probabilities, certainty
factors, or confidence levels and many more techniques to any or
all input data. This feature closely represents how most problems
are handled in the real world. An expert system can take all
relevant factors into account and make a recommendation based
on the best possible solution rather than the only exact solution
to handle such problems. This paper describes the various types
of uncertainty, its sources and different approaches to handle
uncertainty. Uncertainities aries from various resources, and
confounds system designers in a variety of ways. As a result
the method with which a system handles uncertain information
forms is a crucial component of its overall performance
Index TermsUncertainity in Expert System; Expert System,
Fuzzy Logic;

I. I NTRODUCTION
The development of knowledge based systems or computer
programs in which domain knowldege is organised as a
seperated part is a rapidally expanding field in the area of
artificial intelligence. Expert systems, a part of knowledge
based systems provides expertise in the form of diagnosis,
introduction, prediction, advice, consultation and so on. They
may also be used for training purposes for new personnel or to
interpret data or to monitor observations. In other hand, eexpert
system is computer program capable of representing knowledge and reasoning about it in a narrow doomain expertise. Eg.
DENDRAL, modelling the mass spectrometer, uses production
rules. MYCIN, which is used in diagnosis and treatment of
bacterial infection is considered to be first expert system. [1]
In Expert systems the word uncertainty is related to the
working with inexact data, imprecise information, handling
identical situation, reliability of the results etc. An expert
system allows the user to assign probabilities, certainty factors,
or confidence levels and many more techniques to any or all
input data. This feature closely represents how most problems
are handled in the real world. An expert system can take
all relevant factors into account and make a recommendation

based on the best possible solution rather than the only exact
solution to handle such problems. Uncertainty is defined as
the lack of the exact knowledge that would enable us to reach
a perfectly reliable conclusion.
In this paper, we will cover Uncertainity, types of uncertainity and various methods used in resolving those uncertinity in
expert systems.
A. Uncertainity
Uncertainity is essentially lack of information to formulate
a decision. The presence of uncertainity may result in making
poor decision. In our daily life, as human beings, we are
accustomed to dealing with uncertainity - thaats how we
survive. [2]
Classical logic permits form only exact reasoning. It assumes that perfect knowledge always exists and the law of
the excluded middle can always be applied. The general term
uncertainty describes any element of the model that cannot
be asserted with complete confidence. Within this general
condition, there are several distinct types of uncertainty[3]
II. T YPES OF U NCERTAINITY

Imprecision - The information available in the model is

not as specific as it should be. For example, when a
distinct value is required, the information available might
be a range (e.g., the height of plant is between 37 and
43), disjunctive (e.g., the height of plant is either 37 or
43), negative (e.g., height of plant is not 37), or even
unknown (often referred to as incompleteness).
Vagueness - The model includes elements (e.g., predicates
or quantifiers) that are inherently vague; for example,
Plant is in early middle age. A particular formalization
of vagueness is based on the concept of fuzziness.
Inconsistency - The model contains two or more assertions that cannot be true at the same time; for example,
height of plant is between 37 and 43 and the height
of plant is 35.
Ambiguity - Some elements of the model lack complete
semantics, leading to several possible interpretations. For
example, it may not be clear whether stated temperature
is in Fahrenheit or Celsius.

III. S OURCES OF U NCERTAINITY

There are several sources of uncertainity :
Imprecise Language : Our natural language has to be
transposed into IF-THEN rules. but sometimes our language is ambiguous and imprecise.
Data or information can be 1) Incomplete
2) Incorrect
3) Missing
4) Unreliable
5) Imprecise
Uncertain terminology
Uncertain knowledge
Incomplete Information - Information is not sufficient for
the expert system to make decision.
Imprecise data - Different terms are used with the same
meaning or a term has multiple meanings.
Errors :
1) Error related to hypothesis a) Type I Error - Accepting a hypothesis when it
is not true - False positive.
b) Type II Error - Rejecting a hypothesis whrn it is
true - False Negative.
2) Error related to measurement a) Errors of precision
b) Errors of accuracy
c) Unreliability stems from faulty measurement of
data result in erratic data.
d) Random fluctuations
e) Systematic errors result from biasness.
3) Error in induction - Induction proceeds from specific
to general, Expert system may consist of both the
deductives and indeductives rules based on heuristic
informations.
Combination of different expert views : When huge
expert systems require the presence of multiple experts,
there is a low probability that all the experts will lead
to the same conclusion. They might have contradiction
resulting production of conflicting rules.
Thus uncertainity may be induces by the degree of validity
of facts, rule conditions and rules themselves.
When dealing with uncertainity, we should be satisfied just
with a good solution. There are numbers of methods to pick
the best solution in the light of uncertainity. [4]
IV. D EALING WITH U NCERTAINITY
It is very easy to derive decisions for the problems which
are precisely defined and all the specific information is given
or the knowledge is represented with certainty. Various quantitative and qualitative methods have been developed to handle
uncertain or imprecise information in an expert system. Quantitative methods may be classified into one valued approach
based on probability theory, Bayes rules and confirmation
theory, two-valued approach based on Dempster-Shafer theory,

and set-valued approach based on fuzzy set theory, incidence

calculus and rough set theory. Qualitative methods include
modal logics, non monotonicity, plausible reasoning and theory of endorsements.
Obviously, no single method can handle uncertainty perfectly. The basis for the selection depends on the nature
of uncertainty. General methods to be used in handeling
uncertainity in expert systems are Probability based methods which includes :
1) Objective Probability
2) Experimental Probability
3) Subjective Probability
Heuristic methods which includes :
1) Certainity factors
2) Fuzzy logic
Since the boolean approach to reasoning does not solve
the problems in mains involving uncertainity, a number of
theories have been developeed. Some known theories to deal
with uncertainity are [4] :
1) Bayesian Probability
2) Hartley Probability
3) Shannon Theory
4) Dempster Shafer Theory
5) Markov Models
6) Fuzzy Theory
In these theories, a scheme on how to introduce measure which eumerically quantifies uncertainities and how to
propagate and combine there measures of uncertanity during
reasoning is equally proposed.
V. C ONCLUSION
In short many approaches to handle uncertainty in expert
systems were discussed in this paper. All the approaches have
their weakness and strength. We cannot say that an approach is
ideal for any expert systems. It all depends upon the problem
to be solved. In my work Multimedia fuzzy based diagnostic
expert system for pest management in chickpea fuzzy logic
and certainty factor has been used to handle vagueness in the
expert system developed.
R EFERENCES
[1] Jerzy W., Grzymala-Busse, Managing Uncertainity in expert systems,
Proceedings of the 2005 ACM Symposium on Applied Computing, pp.
776-781, 2005.
[2] Bruce Abramson, Keung Chi Ng, Uncertainity Management in Expert Systems, Journal in IEEE Expert: Intelligent Systems and Their
Applications archive. Volume 5 Issue 2, April 1990
[3] Sonal Dubey, R.K.Pandey, S.S. Gautam Dealing with uncertainity in
Expert Systems, International Journal of Soft Computing and Engineering (IJSCE) ISSN: 2231-2307, Volume-4 Issue-3, July 2014.
[4] C. Grosan, A. Abraham Managing Uncertainity in Rule Based Expert
Systems, in Springer-Verlag Intelligent Systems, ISRL 17, pp.187-217
at Heidelberg 2011.