I have 3 variables where 2 are predictors (positive integers) and the response variable is a real number.
To give an example:
| a | b | c |
|---|---|---|
| 1 | 2 | 5 |
| 1 | 5 | 7 |
Because this dataset is unique to my problem, I don't have an equation that can explain the relationship among a, b, c.
However, when I plot these as a surface, there is a nonlinear relationship.
How do I go about to find the maximum value of c while minimizing a, b. I know this sounds like a numerical optimization problem but without an equation, I won't be able to solve it.
Edit: can machine learning help here? Large amount of data can be obtained.
a, ctogether and another one forb, c, or is it one equation for all the variables? How can I generate these equations? – Patrick Jul 28 '21 at 21:09cin terms ofa,b? Because that's easily done using e.g. Gaussian processes; here's a bibliography https://stats.stackexchange.com/a/207025/22311. But finding the maximum ofcis generally different than minimizinga,b. In order to answer this, you'll need to be able to write down the tradeoff that you're willing to accept between largercand smallera,b. – Sycorax Jul 29 '21 at 20:23