In this study, we constructed an olfactory model based on the neural connective structure of the olfactory system and simulated the processing of information in the olfactory bulb. The proposed model simulates selective adaptation, and can predict the perceptual similarity of an odor pair composed of multiple odorants from the neural activity pattern. A strong correlation (r = 0.83, p = 0.04) was confirmed between the predicted perceptual similarities and the discrimination rate obtained from behavioral experiments on mice.