Porous alumina membrane filters (Anodisc(TM)) were applied to catalyst support for oxidative reforming of methane. Two discs were used: the 1st catalyst disc was mainly for methane combustion and the 2nd was for methane reforming. Additives for Ni/Anodisc or Co-Mg/Anodisc catalyst were screened by an artificial neural network (ANN). After an ANN was trained using the physicochemical properties of 9 elements and their catalytic performances as Ni-additive/Anodisc or Co-Mg-additive/Anodisc, the ANN can predict the catalytic performance from physicochemical properties of other elements than the 9 elements. The two kinds of optimum catalysts were determined to be Ni-Pr/Anodisc and Co-Mg-Li/Anodiscs, respectively. Hot-spot formation was avoided according to the restricted catalyst zone by overlapping these quite thin discs. The total performance of the optimized base metal catalysts was almost the same as that of Rh/Anodisc catalyst disc. (C) 2009 Elsevier B.V. All rights reserved.