I have several completely different software components that are producing scores for certain candidate solutions.
Now, setting aside the technical details of what is the problem that I want to solve and how the several software components (riddled with complex heuristics) are obtaining the scores, I would like to know how to combine these (potentially 0..1 normalized) scores.
The goal is to have one final summarizing score, the candidate solution that get the maximum score will be chosen.
One possibility is averaging, another might be multiplying, but I wonder: aren't there any smarter techniques?
The scores might be interpreted as probabilities if they are normalized in the interval [0,1].
Any pointers to literature? Any suggestions/math formulas?