Feature selection is an important technique for identifying informative genes in microarray datasets. In order to select small subset of informative genes from the large datasets various evolutionary methods have been used. However, because of the small number of samples compared to the huge number of genes many of the computational methods face difficulties to select the small subset. This paper proposes a modified PSO algorithm, Minimized Feature Space (MFS) Particle Swarm Optimization to optimize feature selection. In the modified PSO approach we propose a new method which controls a particle's movement towards the best solution. The proposed approach is applied on leukemia, colon and lung cancer benchmark datasets and experimental analysis show good performance.

Feature Selection Particle Swarm Optimization F-score T-score mfspso