Granular Monotonic Binning in SAS

In the post (https://statcompute.wordpress.com/2017/06/15/finer-monotonic-binning-based-on-isotonic-regression), it is shown how to do a finer monotonic binning with isotonic regression in R.

Below is a SAS macro implementing the monotonic binning with the same idea of isotonic regression. This macro is more efficient than the one shown in (https://statcompute.wordpress.com/2012/06/10/a-sas-macro-implementing-monotonic-woe-transformation-in-scorecard-development) without iterative binning and is also able to significantly increase the binning granularity.

%macro monobin(data = , y = , x = );
options mprint mlogic;
data _data_ (keep = _x _y);
  set &data;
  where &y in (0, 1) and &x ~= .;
  _y = &y;
  _x = &x;
run;
proc transreg data = _last_ noprint;
  model identity(_y) = monotone(_x);
  output out = _tmp1 tip = _t;
run;
proc summary data = _last_ nway;
  class _t_x;
  output out = _data_ (drop = _freq_ _type_) mean(_y) = _rate;
run;
proc sort data = _last_;
  by _rate;
run;
data _tmp2;
  set _last_;
  by _rate;
  _idx = _n_;
  if _rate = 0 then _idx = _idx + 1;
  if _rate = 1 then _idx = _idx - 1;
run;
  
proc sql noprint;
create table
  _tmp3 as
select
  a.*,
  b._idx
from
  _tmp1 as a inner join _tmp2 as b
on
  a._t_x = b._t_x;
  
create table
  _tmp4 as
select
  a._idx,
  min(a._x)                                               as _min_x,
  max(a._x)                                               as _max_x,
  sum(a._y)                                               as _bads,
  count(a._y)                                             as _freq,
  mean(a._y)                                              as _rate,
  sum(a._y) / b.bads                                      as _bpct,
  sum(1 - a._y) / (b.freq - b.bads)                       as _gpct,
  log(calculated _bpct / calculated _gpct)                as _woe,
  (calculated _bpct - calculated _gpct) * calculated _woe as _iv
from
  _tmp3 as a, (select count(*) as freq, sum(_y) as bads from _tmp3) as b
group by
  a._idx;
quit;
title "Monotonic WoE Binning for %upcase(%trim(&x))";
proc print data = _last_ label noobs;
  var _min_x _max_x _bads _freq _rate _woe _iv;
  label
    _min_x = "Lower"
    _max_x = "Upper"
    _bads  = "#Bads"
    _freq  = "#Freq"
    _rate  = "BadRate"
    _woe   = "WoE"
    _iv    = "IV";
  sum _bads _freq _iv;
run;
title;
%mend monobin;
Below is the sample output for LTV, showing an identical binning scheme to the one generated by the R isobin() function.

 更多相关文章请看 https://statcompute.wordpress.com/category/scorecard/
分享到:更多 ()