Is there any way to do Operational Research Application (Especially Nonlinear optimization problem)on SAS Enterprise Miner 15.x?
Thanks in Advance!
Hello @C_Golen ,
I do not think so.
I think the Rigorous Constrained Optimized Binning for Credit Scoring is a a mixed-integer nonlinear programming (MINLP) formulation, but for the formulation of your own nonlinear optimization problem(s) you need SAS/OR (Operations Research) or SAS/IML (Interactive Matrix Language).
SAS/OR PROC OPTMODEL has a super-friendly AML (Algebraic Modeling Language) to specify your mathematical program.
Enterprise Miner is not really an option.
Kind regards,
Koen
Hello @C_Golen ,
I do not think so.
I think the Rigorous Constrained Optimized Binning for Credit Scoring is a a mixed-integer nonlinear programming (MINLP) formulation, but for the formulation of your own nonlinear optimization problem(s) you need SAS/OR (Operations Research) or SAS/IML (Interactive Matrix Language).
SAS/OR PROC OPTMODEL has a super-friendly AML (Algebraic Modeling Language) to specify your mathematical program.
Enterprise Miner is not really an option.
Kind regards,
Koen
Hi @sbxkoenk ,
I was thinking the same way. But I wanted to be sure.
Sorry for late replying.
Thanks Again!
Hi @sbxkoenk
I have an extra question due to your answer. Is there any way to add binary variable to MILP? In my situation i have to add a variable about whether weather is rainy or sunny.
Thanks in Advance agian
Hello @C_Golen ,
You can definitely add binary variables to your MILP.
For sure just one, 20 or 50 binary variables may be challenging!
Taken from :
SAS/OR® 15.1 User’s Guide
Mathematical Programming -- The Mixed Integer Linear Programming Solver
(2018)
https://support.sas.com/documentation/onlinedoc/or/151/milpsolver.pdf
[Extract]
It is important to realize that mixed integer linear programs are nondeterministic polynomial-time hard
(NP-hard). Roughly speaking, this means that the effort required to solve a mixed integer linear program
grows exponentially with the size of the problem. For example, a problem with 10 binary variables can
generate in the worst case 2**10 = 1,024 nodes in the branch-and-bound tree. A problem with 20 binary
variables can generate in the worst case 2**20 = 1,048,576 nodes in the branch-and-bound tree. Although it is
unlikely that the branch-and-bound algorithm has to generate every single possible node, the need to explore
even a small fraction of the potential number of nodes for a large problem can be resource-intensive.
Look in this document for 'binary' or 'binary variable' and you will find some example code where binary variables are part of the problem formulation.
Good luck,
Koen