Case-based reasoning: a textbook

Rosina Weber

Outline

Case-based reasoning: a textbook

Abstract

This chapter introduces the basic concepts of Case-Based Reasoning. It does not require any previous knowledge about the topic. The intention is the reader should understand the principal ideas and follow the descriptions of the first examples, the remaining chapters in PART I, and PART II. This chapter provides fundamental basis for understanding the remainder of the book. In most cases no formal definitions are used. In later chapters the concepts will be extended and put on a more formal basis as well as be illustrated by examples.

Figures (9)

In Fig. 2.1 the diagram breaks down the CBR methodology into steps, which helps explain how CBR uses experiences to solve problems. These steps are executed around the concepts of problems and solutions. Problems, solutions, and these steps will cover a major part of this chapter.

In order to make use of experiences for solving the problem of finding a diagnosis and a repair we need a collection of many experiences to choose from. These will be a collection (or set) of cases. As a general advice, we can look at humans describing a failure. That means we did not invent something new here. The CBR goal is just to support the usage of those experiences. Even if all attributes used are of interest, it is not guaranteed that they all have the same importance. This will be considered below when similarity is discussed.

Fig. 2.2 Three types of case organisation: flat, structured, unstructured text We have three main types of case organisation: flat, structured, and unstructured (e.g., text, images). Figure 2.2 illustrates the three basic types of case organisation. Note that these forms already suggest different programming paradigms, but we will only get into the programming aspects in Chap. 5, Case Representations.

Cases are very commonly hidden within texts or images. At this point we will not yet discuss other concerns pertinent to unstructured organisations. We have entire chapters dedicated to them, namely, Chaps. 17, 18, and 19. Table 2.3 Six diagnoses cases

Table 2.4 Attributes in query problem, Case | and Case 2 In order to illustrate how this is done we extend the above example of car faults and look at a query problem and a small case base, containing just two cases, shown in Table 2.4. What we do is to compare the query problem with the problems of the stored cases. This comparison is a crucial step and known as similarity assessment. The goa is to find a case that helps in solving the problem. In other words, the case should be useful for this purpose. The reason is that a case is useful if its problem description is close to the query problem. Similarity is just a word for this. The goal is the cases are analogous in such a A way | that at their solutions can be reciprocally reused. Tli11 sm A A kd oe a a a a

Table 2.5 Comparisons between query problem and cases The second concept within similarity assessment is the relative relevance of each attribute. In practice, each attribute is not equally relevant, and this has to be represented in the similarity assessment. In addition, the problem of describing the importance of the attributes is denoted by a number too. Larger numbers denote a greater importance because more important attributes play a larger role. The comparison is described in Table 2.5 and will later be described in more detail. inde) w@ Oe OS fo eeease ese woccen lepeers. presen aaa A LO EL Ke

Adaptation can be performed on different levels of granularity. One extreme case is reusing the solution strategy. An example is reusing a travel plan. Another extreme case is using the solution itself. Both are called solution adaptation. Cc

The use of cases is a reuse of previous experiences in a new situation. If the new problem situation is exactly like the previous one (which is supposed to have beer successful) then the reuse is simple: Just copy the old solution. The general reuse principle for a selected case is shown in Fig. 2.3. It is rare be able to use a solution exactly as it is recorded. This happens if the new problem situation is not too different in essential aspects from the nearest neighbour selected from the case base. Then the recommendation is to adapt the recorded solution before reusing it to best suit the new problem. This can be done either manually or automatically. CBR presents formal adaptation methods, which we will introduce next.

For the knowledge containers described next we state what kind of knowledge could be contained in them. We say little about how the knowledge is formally represented. In CBR we identify four major knowledge containers. They are presented in Fig. 2.6.

References (8)

Aamodt A, Plaza E (1994) Case-based reasoning: foundational issues, methodological variations, and system approaches. AI Communications 7 (1):39-59
Bridge DG (2005) The virtue of reward: performance, reinforcement and discovery in case-based reasoning. Keynote at ICCBR 2005: 6th international conference on case-based reasoning, Chicago, IL, USA, August 2005. Awards, Honours, Affiliations. http://www.cs.ucc.ie/~dgb/recognition.html Accessed 28 Feb 2013
Carbonell JG (1983) Learning by analogy: formulating and generalizing plans from past experience. In: Michalski R, Carbonell JG, Mitchell T (eds) Machine learning: an artificial intelligence approach. Springer-Heidelberg, p 137-159
Cbrwiki (2011) Case-based reasoning Wiki. http://cbrwiki.fdi.ucm.es/wiki/index.php. Accessed 18 Jul 2011
Kolodner JL (1993) Case-based reasoning. Morgan Kaufmann, San Mateo, CA Recio-García JA, González-Calero PA, Díaz-Agudo B (2013) jColibri2: A framework for building Case-based reasoning systems. Science of Computer Programming (in press)
Richter MM (1998) Introduction. In: Lenz M, Bartsch-Spörl B, Burkhard H-D et al (eds) Case-based reasoning technology: from foundations to applications. Lecture notes in artificial intelligence, vol 1400. Springer, Berlin, p 1
Richter MM, Aamodt A (2005) Case-based reasoning foundations. The Knowledge Engineering Review 20(3):1-4
Schumacher J (2002) Empolis Orenge -an open platform for knowledge management applications. In: Minor M, Staab S (eds) Experience management: sharing experiences about sharing the experience. Papers from the 1st German workshop on experience management, Berlin, March 7-8, 2002. GI, Bonn, p 61

Case-based reasoning: a textbook

Sign up for access to the world's latest research

Abstract

Related papers

References (8)

Related papers