Information Technology Reference
In-Depth Information
Fig. 5.6. Central fishing grounds after clustering and human revision
With the help of tools we have developed, domain experts can classify the
historical data in a short period of time.
5.11.2
Similarity Measurement
A key point is CBR is how to retrieve the “best” similar cases. Case retrieval
is based on the similarity compares, and totally determined by their definition
of “similarity”. It is hardly the case that we can find satisfactory cases with an
un-proper definition of similarity. So the similarity definition carries a great
weight in CBR systems. However, similarity is a notion that is always coarse,
incomplete, and ever-changing. The optimum choice of cases usually follows
the following rules: Compute the semantic, structure, target, and individual
similarities using the corresponding algorithms to assess the overall
similarities between cases; then calculate the whole similarity according to
the importance of these similarities in the question, it is similar to the total
dynamic weighting.
In our system, we employ the following three methods to measure the
similarities:
·Similarity in terms of position
