5

I've never done anything with statistics, so I'm unsure of how to start with this problem that I need to solve for some software that I'm working on.

Basically, there are three parameters

  1. An integer parameter (A) that can be a range of (for the sake of argument) 1-10
  2. A Boolean parameter (B) that if present weights the value of A.
  3. An integer parameter (C) that can be a range of 1-100.

There is one outcome: An integer (D) that can be of the range -1, 0, 1.

For each value of C, there will be one value of A, B, and D. C is a preexisting condition. A, and B are parameters. D is the outcome, which may be subjective.

+---+----+---+----+
| C | A  | B | D  |
+---+----+---+----+
| 1 |  3 | 0 | -1 |
| 1 |  3 | 1 | -1 |
| 1 |  5 | 0 |  0 |
| 1 |  5 | 0 |  0 |
| 1 | 10 | 0 |  1 |
| 1 | 10 | 1 |  1 |
| 2 |  3 | 0 | -1 |
| 2 |  3 | 1 |  0 |
| 2 |  4 | 0 |  0 |
| 2 |  4 | 1 |  0 |
+---+----+---+----+

What I'm looking for, is given C, A, B, how can I make a prediction of what D will be. In this case, I will have historical data of all four values to go from. C won't necessarily be linear as in my example.

  • This seems like a good fit for an Adaptive Logic Network. – Robert Harvey Sep 28 '15 at 5:57
  • My statistics may be a little rusty, but I'm not sure what D is supposed to represent. Looks like it could be a sort of coefficient of determinination, correct me if I'm wrong. – Neil Sep 28 '15 at 6:29
  • On a side note, with 100 possible values for C * 2 possible values for B * 10 possible values for A, you could code this up as a lookup table pretty easily. – Dan Pichelman Sep 28 '15 at 13:43
4

This is a supervised learning problem.

It seems to be a classification task i.e. the output variable D takes class labels (the other group is the regression task).

There are a lot of algorithms for classification and you should probably start with something simple e.g. logistic regression.

If I misunderstood the example and yours is a symbolic regression task, a good approach is genetic programming (you have to specify a set of primitive and new equations are then formed by recombining previous equations).

  • You've given me enough to start investigating. I'm in completely new territory, so I didn't even know the right terminology. Thank you very much! – kettch Sep 28 '15 at 18:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.