1
$\begingroup$

I am trying to predict results of football games. Some of our attributes only give meaning for a prediction only when they are considered in interaction with another attribute. To illustrate, a team with an average possession of 48.6 has much higher possession than some teams and much lower than others. Should we modify the features in such a way that we would use the difference between the two teams, instead of two values? On the other hand some algorithms might be good at finding this interactions?

$\endgroup$
0
$\begingroup$

Should we modify the features in such a way that we would use the difference between the two teams, instead of two values?

Yes, you should definitely design your features "to help" the ML algorithm, this is called feature engineering. Be careful though that there can be many different options for representing the contrast in "average possession": you can think of the difference indeed, but also a ratio or many other arithmetic combinations.

On the other hand some algorithms might be good at finding this interactions?

Most ML handle take into account interactions between features to some extent, but it depends what kind of interaction and which ML algorithm.

$\endgroup$

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.