In　order　to　overcome　the　defects　of　slow　convergence　speed　and　low　precision　appeared　in　the　original　artificial　bee　colony　(ABC)　algorithm,　a　novel　and　improved　adaptive　ABC　algorithm　is　presented　in　this　paper.　By　dynamically　adapting　the　step　length　that　controls　the　range　of　neighborhood　during　the　process　of　search,　the　proposed　algorithm　produces　three　candidate　solutions　that　have　good　performances　in　exploiting　in　small　search　space,　exploring　in　large　search　space　and　remaining　initial　search　space,　respectively.　For　illustration,　a　single　variable　function　is　utilized　to　analyze　the　cause　of　low　precision　and　slow　convergence　speed.　In　addition,　a　different　probability　selection　strategy　is　introduced　to　maintain　population　diversity　of　the　bee　colony.　The　improved　ABC　algorithm　is　tested　on　five　numerical　optimization　functions　and　compared　with　the　original　ABC　algorithm　and　a　novel　ABC　algorithm　known　as　ABC-SAM.　The　results　show　that　the　improved　ABC　algorithm　is　superior　to　two　other　algorithms　on　convergence　and　optimization　precision.