Data　streams　scheduling　is　a　typical　NP-complete　problem.　To　solve　the　problem　of　data　streams　scheduling　on　heterogeneous　cluster　effectively,　a　self-adaptive　strategy　based　on　improved　particle　swarm　optimization　is　proposed.　Inspired　by　the　gene　theory,　a　transgenic　operator　is　designed　to　keep　the　best　individual　and　improve　the　ability　of　local　solution.　The　mutation　operator　is　built　into　the　proposed　strategy　to　maintain　population　diversity　and　improve　the　ability　of　global　exploration.　Simulation　results　show　that　the　proposed　strategy　gives　a　good　balance　between　local　solution　and　global　exploration　and　has　excellent　efficiency　in　data　streams　scheduling.