Oxidative reforming of methane was conducted at 650 degrees C, 1 MPa, and GHSV= 2,000,000 N ml/(g h) using Ni/alpha-Al2O3 catalyst. Dilution of catalyst bed with alpha-Al2O3 prevented the formation of hot-spots in the catalyst bed. Element X (X = B, P, Ca, Mn, Fe, Cd, Ce, Gd and Re) was added to Ni/alpha-Al2O3, and the catalyst activities were experimentally observed to obtain training data of an artificial neural network (ANN). Then the physicochemical properties of element X and the observed values (CH4 conversion, H-2 selectivity, or CO selectivity) of Ni-X/alpha-Al2O3 catalyst were used for ANN training. After the training, the ANN was able to predict the catalytic performance of Ni-Z/alpha-Al2O3 based on the physicochemical properties of element Z where Z is a possible additive other than X. In addition to La and Ce, Sc and Nd were predicted to promote the activity of Ni/alpha-Al2O3. The experimentally observed activity of Ni-Sc/alpha-Al2O3 Was five times higher than that of unpromoted Ni/alpha-Al2O3 catalyst. (C) 2008 Elsevier B.V. All rights reserved.