How to implement Evolutionary Solver xls in SAS

[Alexa_Rau]

Hello guys,

I need your help in implementing the Evolutionary Solver from Excel (or similar) in SAS.

I have to optimize the confusion matrix and Matthew correlation coeff.

In my table I have the predicted condition which is affected by the solver condition.

My table

A           B           ln(A)    ln(B)     Predicted_cond  Observed_Condition

1.02      0.01       0.02    -4.61         1                      1

1.03      0.01      0.03     -4.61         1                      0

1.01      0.001      0.01     -6.91         0                    0

1.01      0.001      0.01     -6.91         0                    1

1.13      0.001      0.12     -6.91         0                    0

1           0.001      0           -6.91         0                    0

1.37       0.01      0.31     -6.91        1                    0

1.16      0.001     0.15     -6.91         0                    1

predicted_cond=1 if A>exp(avg(alpha)) and B>exp(avg(beta)) 

Based on predicted and observed condition Matthew Correlation Coefficient (MCC) is computed.

My objective is to maximize MCC, taking in account that predicted condition is dynamic.

Values to be change in order to maximize the MCC are exp(avg(alpha)) and exp(avg(beta)) .

 I saw that proc GA has the capability to solve this, but I don't figure out how to implement it.

Thanks,

[Rick_SAS]

Since you mention PROC GA, which implements a genetic algorithm, I will mention that PROC IML also has an implementation of a genetic algorithm. There are many examples in the doc.