Mutation　is　one　of　important　operator　in　genetic　algorithm,　but　the　studies　we　have　performed　showed　that　conventional　mutation　based　on　floating　coding　had　some　shortcoming　such　as　random　direction,　lacking　self-study　and　self-adaptation,　greedy　local-search　and　so　on.　Aimed　at　foregoing　question,　BP　mutation　algorithm　that　characterized　by　well-known　error　back　propagation　rule　was　presented　in　this　paper.　This　mutation　algorithm　can　mutate　on　the　basis　of　individual　fitness,　so　it　is　guiding-fine,　self-study,　self-adaptation.　The　results　of　the　compare　indicate　that　the　proposed　algorithm　improve　greatly　efficiency　and　quality　of　genetic　algorithm.