Feature Selection Using Intensified Tabu Search for Supervised Classification
Muhammad Atif, Jim Smith · 2008
Feature selection (FS) algorithms are popular methods to reduce the dimensionality of the feature space and remove the redundant, irrelevant or noisy data and improves the classification accuracy. In this chapter; we have dis- cussed our recently proposed tabu search based algorithms for feature se- lection problems and have compared with other sequential feature selection algorithms. Tabu search is refined with respect to traditional approaches: A feature search intensification strategy is introduced. Search intensification is realized by incorporation of intermediate term memory in the search process. We have also discussed advanced tabu search approaches for supervised classification. Like many combinatorial optimization problems such as VLSI, Time Tabling, and Scheduling in which TS is quite useful, TS is also quite effective in data mining feature selection problems.