cancel
Showing results for 
Search instead for 
Did you mean: 

Reverse engineer a formula for a Solver like question - can I use Optimize tool?

florini
Meteoroid

Hi,

The case is as following:
- I train one past data using the logistic formula to estimate a probability percentage.
- I apply the formula on the future data and associate a % with each record.

In the logistic formula there is a factor that can be influenced - and I need to apply a prescriptive column on it.
Question:

- how to calculate the recommended factor value to maximize the % for the future data records?

I see 2 solutions:

A. To reverse engineer the formula, so if I input the desired %, I get the factor value?

B. To use Optimize Tool - for this I would need some help, and see if it really meant for this.
b1. is it approapite for detecting one value in a maximization problem, or is only working with pairs of variables to be solved?
b2. Can it handle a formula of a logarithmic regression (see below)?

Many thanks,
F


Formula:

(EXP(-1.5173786-0.0007703*[Variable1]+1.7769691*[Variable2]-0.0077538*[Variable3]+0.0479129*[Variable 4]))/(1+(EXP(-1.5173786-0.0007703*[Variable1]+1.7769691*[Variable2]-0.0077538*[Variable3]+0.0479129*[Variable 4]))) = Y

I need to find out (for each line) what is the value for Variable 4 for a given Y

Pulsar
Pulsar

I  wonder if either of the following topics might lend insight into finding a solution:

 

Replicating Goal Seek functionality (or) Optimization with Dynamic Constraints

 

In both cases, iterative macros were utilized, and R is always a possibility also, (mentioned