Case-based tool for treatment of behavioral problems

In case-based reasoning (CBR) systems, solutions to new problems are found by retrieving and modifying the solutions to similar old problems. Many CBR systems use a two step approach for retrieving a similar past case. This paper describes an enhanced approach, which was developed for a CBR decision support tool for treating behavioral problems in Alzheimer's disease patients. Here, the new case is first screened to determine whether or not it should be compared at all. If the new case needs to be compared with the old cases, then an expanding window initial discrimination is used to find potential cases in the case library. The next step is to apply a similarity metric to these potential cases to find the most similar case. The previously stored classification of the most similar case is then extended to the new case. Once the new case is classified, a rule-based approach is used to propose a solution based on the classification and the features of the new case.