CO2 emissions reduction has become a hot issue to relieve global warming in recent years. CO2 absorption with amines aqueous solutions is a promising method for its capture, and increasing CO2 capacity and reducing energy consumption are essential for its application. In this work, a data set of 29 different amines was regressed with genetic function approximation (GFA) and artificial neural network (ANN) algorithm to find a predictive quantitative relationship between amine structure and CO2 absorption capacity. Density functional theory (DFT) methods at levels B3LYP and 6-311+g (d,p) were used to optimize the structures. Both reactants and products were introduced in regression, and an optimal model with best relevance and predictability for CO2 absorption at 313 K was obtained and experimentally confirmed. Descriptors analysis showed that reducing the number of hydroxyl groups, increasing molecular mass, and changing steric effect of amine structure would increase its CO2 absorption capacity. Two new amine structures predicted with high absorption capacity were proposed.