Grid research and engineering require behavioral models of its use and middleware. These models must be inferred from the data, through Machine Learning. Technology and usages of Grids and Clouds change rapidly, challenging analysis and evaluation. We summarize two of these challenges and present new methods for addressing them.