0

Current state: I have a dataframe with input states and results of the decision made. Currently, the decision is made based on the binary state (true/false) of the variable.

Goal: Since each variable has a different weight on the outcome, we need to find the optimal weights that can replace the true/false condition to improve the probability score.

assumptions are

  1. Not all variables have strong correlation to end result (using pearson method).
  2. There exists a set of weight given state of two primary variables (var0, var1) that can improve prediction by few percent(at least better than hardcoded numbers).

I have looked at RL but that makes the process and computation too complex to start from beginning(I have just started this path) given number of variables.

Is there algorithm (M.L or otherwise) or method that can formulate the weights of given variables?

Any recommendations / knowledge sharing?

Dataframe looks like this,

result1 result2  var0   var1 var2  ..  var8
-3.79       8    -9.8     4    3      -34.61 
 5.16       11   44       3    10     -80.81
-3.22       5    52.58    5    10     -13.5   
-0.07       7    124.48   6    0      -75.14   
13.75       32   44.76    8    8      -72.33
desertnaut
  • 57,590
  • 26
  • 140
  • 166
stack
  • 51
  • 4
  • 1
    Sounds like a good use case for (multivariate) linear regression https://en.wikipedia.org/wiki/Linear_regression – g_uint Nov 04 '22 at 09:43

0 Answers0