Porous alumina ceramics are fabricated using the sacrificial template method to investigate the impact of two different sphere sizes, PMMA porogen (<i>D</i>50 of 80 and 163µm) and volume fraction (up to 76 vol.%) on material properties. In order to screen the viable parameter space (where samples are mechanically resistant and permeable) efficiently and to leverage the possibility to manufacture several samples in parallel, a batch active learning algorithm coupled with a data-driven model based on Gaussian process classification is developed. Complementary, a Gaussian process regression (GPR) model is developed to predict permeability trends across the viable parametric space. Both models are tested and blind-validated experimentally using unseen data. The combination of GPC and active learning offers a powerful tool for generating effective experimental plans, particularly when investigating processes with large parametric spaces.