The　purpose　is　to　apply　the　Binary　Particle　Swarm　Optimization　(BPSO)　in　feature　selection.　According　to　feature　selection,　the　method　of　particles　coding,　fitness　functions,　and　feature　selection　functions　were　designed.　Furthermore,　trans-gene　operator　was　adopted　to　solve　BPSO＇s　premature　convergence.　The　simulation　experiment　results　show　that　the　feature　subsets　selected　by　this　new　algorithm　are　representative.　The　conclusion　is　that　this　algorithm　is　available　and　feasible　in　feature　selection.　©　2008　IEEE.