The document proposes a Complementary Particle Swarm Optimization (CPSO) method for feature selection in DNA microarray data. CPSO was designed to overcome limitations of standard PSO getting trapped in local optima. CPSO uses a complementary strategy to move particles to new search regions. It was tested on six microarray datasets and achieved lower classification errors than other methods. Future work will combine CPSO with K-Nearest Neighbors classification to potentially further improve performance.