Ask HN: Best way to group data using machine learning without clustering?

I want to create group of matches(1-1, 1-2, 1-3 or m-m) from the data. I already have test data. Which algorithm should I use? Clustering is not at all helpful!