Dropbased latches (English case based reasoning, franz. raisonnement par cas) is a machine learning procedure for problem solution by analogy reasoning. The central element in a CBR system is a drop basis in such a way specified (drop data base, case MEMORY), in which problems already solved are stored as case. Such a case consists at least of a problem description and an associated problem solution. The goal is to consult for the solution of a given problem the solution of a similar and in former times already solved problem. Thus one copies a human behavior: Before a new problem posed, humans often remember a comparable situation, which he experienced in the past, and try, to master the current task similarly.

Occasionally one speaks also of memory-based reasoning.

Proceeding

The probably most well-known model decreases/goes back to the scientists Agnar Aamodt and Enric Plaza, which the basic principle of the Case Based Reasoning as a process with four phases, which described CBR cycle so mentioned, (source: see below).

1. Retrieve. On the basis of a given problem description it applies to determine in the drop basis as similar a problem as possible. The challenge in this phase consists of determining the similarity of the problem descriptions.
2. Staggered array. The solution of the case, which is most similar to the given, is taken over as a first proposal for solution. Thus one has a starting point for the solution of the new problem.
3. Revise. Always one cannot solve the current problem exactly the same like the earlier. In the Revise phase one examines the initial solution won before and adapts her if necessary to the concrete conditions.
4. Retain. The revised case is stored finally in the drop basis and is available thereby for future inquiries. In this way the system with each further solved problem learns in addition and improves so its efficiency.

Application

Case Based Reasoning worked satisfactorily particularly in application systems for the customer service, Help Desk systems so mentioned, where one uses it e.g. to the diagnosis and therapy of customer problems. In recent time one uses it strengthened in (product) consulting systems, for example in the E-Commerce, as well as for the classification of texts.

As favourable it is considered that CBR also with badly structured and incompletely described problems can be used. Contrary to neighbouring concepts (see below) is initially already sufficient a comparatively small collection of cases of reference, which increases by the work with the CBR system gradually. Also in application domains, whose admits exact effect connections not completely are, CBR is suitable.

Like always, if one argues with analogies, it is to be made certain that the proposals for solution for the available problem, generated by the system, are adequate, whether thus for example the conditions, on which the historical solution was based are still fulfilled etc. (obsolescence of the knowledge).

Classification

The Case Based Reasoning is a subsection of artificial intelligence and can herein to the machine learning procedures be counted. The learning process is based on analogy, in contrast to learning by induction and deduction. Due to numerous application possibilities in enterprises (see above) one is occupied not only in (core) computer science, but also in the economical informatics with CBR.

Literature

• Aamodt, Agnar; Plaza, Enric: Case Based Reasoning: Foundational Issues, Methodological variation, and system Approaches, AICOM 7 (1994) 1, P. 39-59; Pdf version.
• Judge, Michael M.: Dropbased reasoning. In: Gnther; Rollinger, Claus Rainer; Schneeberger, Josef (Hrsg.): Manual of artificial intelligence. 4. Edition, Munich/Vienna 2003, P. 407-430. ISBN 3486272128.
• Miner, Ralph: Experience management: Foundations, development Methodology, and Internet Based Applications, LNAI 2432, Springer (2002), ISBN 3-540-44191-3

Related links

Case based Reasoning website OF the IIIA

Articles in category "Dropbased reasoning"

We found here 11 articles.