Multidimensional optimisation of a collected 4D dataset

Hi all,

I have collected some data where I manipulate 2 variables (X, and Y) and observe 2 outcomes (A and B).

I now want to find the solution which minimises X and A whilst maximising Y and B.

From searching the Matlab help and forums this seems like a multidimensional optimisation problem, and a colleague of mine actually suggested using a pareto front, which seems like the right track. But, I'm not really sure where to start.

One problem I have come across is that a lot of functions require a 'model' input, which seems to be a mathematical description of the dataset to optimise. But I don't have that for my dataset because it has been collected from real interactions. Does this mean I have to fit an n-d function to the data first?

Thanks for any help,

M.