Computational Intelligence Techniques for Classification in Microarray Analysis - Computational Intelligence in Healthcare: Advanced Methodologies - page 282

Fig. 1. Structure of the proposed CBR model
3.1 Retrieve
Traditionally, only the cases similar to the current problem are recovered, often be-
cause of performance, and then adapted. With regards to expression array, the number
of cases is not a critical factor, rather the number of variables. For this reason, we
have incorporated an innovative filtering strategy where variables are retrieved at this
stage and then, depending on the identified variables, the rest of the stages of the CBR
are carried out. The new strategy allows a notable reduction in the dimensionality of
the data.
3.2 Reuse
In this phase the clustering of individuals is carried out, along with the classification
of new individuals to one of the clusters. This chapter proposes a hybrid solution that
takes into account a ESOINN neural network and the PAM method.
3.3 Revise
As shown in Figure 1, the revision is carried out by a human expert who determines if
the group assigned by the system is correct. To facilitate the human expert task, the
equivalence index and the error rate were calculated in the reuse stage. It is important
for the medical human expert to understand the classification process performed in