Information Technology Reference
In-Depth Information
target case related to those similar cases can be calculated out. The retrieval
algorithms rely on the indices and the target case to search in the case base to
find potentially useful cases. The algorithms need the ability to perform
partial matches between the target case and cases in the base. Then the system
retains a group of partial matches with potential use. Finally the system ranks
these partial matches and chooses the most useful case.
There are three key points in the case retrieval: The retrieval algorithms,
matching function, and situation assessment. The following discussion
focuses on the retrieval algorithms. Algorithms are closely linked to the data
structures that they process. So we can not isolate our discussion on the
retrieval algorithms to the organization of case bases. Bases with different
organization should be searched by different algorithms accordingly.
Different organizational forms have both pros and cons, and the choice of
them should be determined by the application in use.
Retrieval
Retainion
Analyze the scene; Analyze the scene;
Refine the description of target case; Refine the description of target case;
Determine the indices of target case; Determine the indices of target case
Search in the well-organized Search in the well-organized
case base and get some partial matches; case base and get some partial matches
Partial Matches Add the new case to the base
Finally select a best match
Fig. 5.4. The process of retrieval and retainion
There have been formed a series of methods for cases' organization and
retrieval in CBR: the serial or parallel ones; the flat or layered ones; the ones
with indices on the small or the great granularities to distinguish different
examples. The mostly-used methods are of based on the inverted index, which
can search in the case base serially or parallelly. The application system
constructed by Kitano can already deal with more than 25,000 cases. Among
