Currently, there are checks in place that limit the use of split_digits to the range [0,8] in the OptimalBinning classes.
Is there any reason to not allow for negative numbers here, to allow for rounding to 10s/100s? The numpy.round implementation that is used will happily accept negative numbers to allow for rounding to powers of 10, e.g. numpy.round(1343.21,-2) == 1300.0.
As for the work required, if my understanding is correct the only thing that would need to be changed to allow for this is to adjust the check 0 <= split_digits <= 8 that is present in several parts of the code.
(N.B. this is the first issue I have ever raised, please let me know if this is the correct way of doing so)
Currently, there are checks in place that limit the use of
split_digitsto the range [0,8] in the OptimalBinning classes.Is there any reason to not allow for negative numbers here, to allow for rounding to 10s/100s? The
numpy.roundimplementation that is used will happily accept negative numbers to allow for rounding to powers of 10, e.g.numpy.round(1343.21,-2) == 1300.0.As for the work required, if my understanding is correct the only thing that would need to be changed to allow for this is to adjust the check
0 <= split_digits <= 8that is present in several parts of the code.(N.B. this is the first issue I have ever raised, please let me know if this is the correct way of doing so)