BACKGROUND: Culture medium is a key element to be defined when biotechnologies are chosen for agro-industrial wastes reutilization. This work aimed at definition of culture medium composition using four agro-industrial wastes (wheat bran, type II wheat flour, soybean meal and sugarcane bagasse) in solid-state fermentation (SSF) of Rhizopus oligosporus, for high-level production of amylases through approaches based on artificial intelligence (AI) or response surface methodologies (RSM). First, substrates were individually assessed. Then, I-optimal mixture experimental designs were performed to determine the influence of two sets of ternary agro-industrial waste mixtures on amylase and specific amylase activities. RESULTS: The best individual substrate for amylases production was wheat bran (392.5 U g(-1)). As a rule, no significant interactions among substrates affecting amylase activities were observed for ternary systems and the approaches under consideration. A significant exception was the amylolytic activity for mixtures composed of wheat bran (91% w/w) and soybean meal (9% w/w). This finding was confirmed analytically by a combination of artificial neural network (ANN) and genetic algorithm (GA). The AI approach improved modelling quality with respect to RSM for production of fungal amylases in SSF. CONCLUSION: The I-optimal design in conjunction with ANN-GA is suggested to optimize accurately culture medium to maximize amylase production by SSF. (C) 2016 Society of Chemical Industry