Case-based reasoning techniques for estimation

Estimation often involves reasoning by analogy. In property valuation, for example, experts are guided by previous similar transactions, using their judgement to adapt the estimated value of a property according to its individual characteristics. Existing CBR approaches, however, rely on domain-specific rules to adapt a similar case retrieved from case memory to match the target case. The expertise necessary for case adaptation may be difficult to capture in the form of rules. The author describes an approach to case adaptation which does not rely on domain-specific rules. The relative importance of case attributes which influence the value to be estimated (e.g. price) guides the retrieval of existing cases which are similar to the target case. The aim, however, is not to find a single most similar case as in other CBR approaches. Instead, the aim is to find a previous similar case whose value will provide an upper bound for the value of the target case and another whose value will provide a lower bound. An overall ranking of the target case in comparison with the two similar cases provides a basis for estimation of the value of the target case by interpolation between its upper and lower bounds. The techniques described should be applicable to a wide range of estimation tasks and have been implemented in a prototype CBR program in the domain of residential property valuation