This is my problem:

I have 10 variables that I intend to evaluate two by two (in pairs). I want to know which variables have the strongest relationships with each other. And I'm only interested in evaluating relationships two by two. Well, one suggestion would be to calculate the pairwise correlation coefficient of these variables. And then list the pairs with the highest correlation coefficient to the lowest correlation. That way I would have a ranking between the most correlated to the lowest correlated pairs.

My question is: Is there anything analogous in the world of artificial intelligence to the correlation coefficient calculation? That is, what tools can the world of AI / Machine Learning offer me to extract this kind of information? So that in the end I can have something like a ranking among the most "correlated" pairs from the point of view of AI / Machine Learning?

In other words, how do I know which variable among these 10 best "relates" (or "correlates") with variable 7, for example?


It sounds like you have a series of data points, each with 10 related measurements, and you want automatically assess which of the measurements are most closely related to each other.

You are right that the correlation coefficient is a good choice for this.

Other techniques used in some AI algorithms include the Information Gain measurement (where you measure the reduction in entropy of one variable that follows from partitioning on another one), and embedded feature selection approaches, like the one in this paper.


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.