Question

Knowledge application systems are typically enabled by intelligent technologies.

Explore Artificial intelligence as a technology for applying knowledge by explaining the differences between rule-based and case-based reasoning systems?

Answer 1

Comparison of Figures 1 and 2 provides a useful perspective on RBR and CBR. The general solution structures of both RBR and CBR are quite similar: in both some matching (against either rules or cases) is performed and then some procedures (inference mechanisms or adaptation techniques) are used iteratively to generate the desired solution. This hints to the fact that it may be possible to seamlessly integrate RBR and CBR; however, there are some important differences between RBR and CBR, as typically discussed in the literature. The contents of the case library are, in general, more complex structures than simple IF. . . THEN.. . rules, and a variety of representational schemes (such as frames and memory organization packets (MOPS) [5]) have been used in the literature to represent their complexity. Pattern matching in CBR is usually more involved than simply matching the left-hand side (LHS) (antecedents) or the right-hand side (RHS) (consequents) of rules (in RBR). This complexity is caused by the data structures used to represent the case library and the fact that the match of the input problem with the various cases in the case library is

usually partial. Unlike RBR, where a solution is obtained by incrementally searching the rule graph, CBR typically generates a complete solution (the selected case from the case library) first, and then progressively adapts it to solve the problem at hand. Although there are well-defined procedures (such as backward or forward chaining) for searching a rule graph, there are no equally well-defined procedures for case adaptation as it is usually ill structured and thus less susceptible to generalization.

• case-based reasoning will not try to induce a generic model from your past observations/experiences, but will simply try to predict/find the answer of a new problem by looking at past experiences. This means you have to maintain a memory of past experience, and then select the most adequate one to answer your new problem. The most well-known representative of such approaches is probably K-NN.

• rule-based approaches will try to induce a "compact" model from your past observations/experiences, and to generalise (make an induction) from those. Once this is done, you can forget your past experiences and only use your learned model to make prediction.