Learning ON Large Datasets Using Bit-String Trees