Abstract:

This paper addresses the elusive quest for that one single best method for data integration. The authors assert that this is a fool's quest since at the heart of learning theory is the famous No Free Lunch Theorem which makes this an impossible mission. The results from four different projects are shown, each one being a genuine real-life commercial problem, in which a number of standard data integration methods were applied. None of these methods is the best in all applications. For any specific problem, the best approach is to find the method that is crafted according to the exact circumstances.