User`s guide

9031666 E9 What is SpectroRx?
1-5
What Is Case-Based Reasoning?
What Is Case-Based Reasoning?
Case-based reasoning systems try to approximate the problem-solving process
used by the human mind. For example, if you are confronted with a new
problem, you analyze past experiences that are similar to the present one.
Then you select a solution that worked for one of these experiences and adapt
it to the current problem. In a sense, your new solution is a gamble or a good
guess because you do not know if it will really work. However, regardless of
the success of the solution, the experience provides valuable new information
that you can refer to in the future. As you tackle more problems, each
experience makes you more knowledgeable. Your first solutions are based on
good guesses but, as time passes, they are based on increasing expertise.
Cased-based reasoning systems work in a similar way.
By definition, a case-based reasoning system must be able to:
• Learn from experience
• Offer solutions to new problems based on past experience
A case-based reasoning system relies on a database of cases just as people rely
on their experiences. Case-based reasoning systems use a retrieval algorithm
to evaluate the cases in the case library and then find the cases that are
similar to the current problem. As cases are added to the case library, the
system is more likely to find cases that are very similar to the problem. The
case-based system shows you how similar problems were resolved so that you
can adapt the various solutions to your needs. Furthermore, the case library
maintains a stable and accurate record of experience that is not affected by
employee turnover.
How Case-Based Reasoning Systems Differ from
Expert Systems
Case-based reasoning systems should not be confused with expert systems or
rule-based, problem-solving systems. Expert systems hardcode a specific
solution to a specific problem. For example, if Problem 1 occurs, then the
expert system recommends Solution 1. The difficulty occurs when your
problem has many similarities to Problem 1 but differs in a few ways. In this
situation, the rule-based system cannot help you even though its database, or
even Solution 1, contains a good solution. Expert systems are sometimes
termed as “brittle” because they break down when their rules do not apply.
They do not have the ability to manage new or ambiguous situations. For
example, they cannot evaluate problems to detect their degree of similarity to
other problems. Rule-based systems work for fixed, static environments but
are inadequate for the fluid, rapidly changing world of information networks.