1
$\begingroup$

I am looking to develop a simple algorithm that is given a set of 10 vectors for a year from which to predict the "winner". This will be a supervised task where the algorithm will be trained with 20 years of data and thus will know the 20 winners.

My question is: what topic or style of machine learning algorithm is this? I am new to machine learning and planning to research more on this style/task before I begin to develop my algorithm.

$\endgroup$

1 Answer 1

0
$\begingroup$

There are different options for the design of this task as a ML problem. A common approach would be to treat this as a regression problem, where the goal is to learn a score for each of the 10 "vectors" (features). The vector with the highest score is the "winner".

Ideally there is some real score value for each vector which can be used as target variable (it must satisfy the "highest score = winner" condition). But if not it's not a big issue: for the training, the score can be 1 for the winner and 0 for the 9 other vectors. This way the model should be able to learn the distinction winner/loser and to assign a score representing the likelihood of a vector to be the winner.

Mind that this is a quite advanced kind of problem imho, it might be a bit difficult for a beginner.

$\endgroup$
1
  • $\begingroup$ Hi! Yes I think I'm going to approach this as a "binary" classification problem, where the classifications are winner and loser. $\endgroup$
    – anas
    Dec 7, 2021 at 1:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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