Graphics Reference
In-Depth Information
Fig. 7.1
Search space for FS
More realistic problems in DM do not only have three features, thus the search
should not start with the full set of features. A search direction must be specified and
different search strategies should be adopted to try to achieve optimal subsets in real
problems.
7.2.1.1 Search Directions
With no prior knowledge about the problem, the search for a optimal subset can
be achieved from the empty set, by inserting new features, or from the full set, by
removing features, with the same probability. These two are the directions of search:
•
Sequential Forward Generation (SFG)
: It starts with an empty set of features
S
.
As the search starts, features are added into
S
according to some criterion that
distinguish the best feature from the others.
S
grows until it reaches a full set
of original features. The stopping criteria can be a threshold for the number of
relevant features
m
or simply the generation of all possible subsets in brute force
mode.
•
Sequential Backward Generation (SBG)
: It starts with a full set of features and,
iteratively, they are removed one at a time. Here, the criterion must point out the
worst or least important feature. By the end, the subset is only composed of a
unique feature, which is considered to be the most informative of the whole set.
As in the previous case, different stopping criteria can be used.
