The search for feature selection methods has been increasingly present in machine learning
applications, especially in those where the number of available attributes is in the range
of hundreds or even thousands. Such applications include, for example, word document
processing, gene expression analysis, and combinatorial chemistry. Feature selection or
selection of characteristics is a concept that proposes methods that aim to provide faster
and more economical predictors, improve predictor prediction performance, and provide
a better understanding of the underlying process that generated the data. Mathematically, feature selection is formulated as a combinatorial optimization problem. In general,
addressing such problems in a way that finding the exact solution is not always feasible.
In this way, computational intelligence methods can be used to allow the feature selection
in practice. Therefore, the objective of this work is to present and propose optimization techniques guided by strategies of feature selection, among which we can highlight
the optimization by swarm of particles, optimization of swarm by competition and the
combination of both.
|
