In this paper, we propose a Kansei Retrieval model using a neural network. It is very difficult to design Kansei Retrieval Agent to imitate a user's Kansei. To solve this problem, we use the neural network, which is comparatively easy to design. The neural network is a mathematical model that imitates brain functions. Therefore, we assume that the input of the neural network is a stimulation that a user received, and the output of the neural network is a user's impression of the stimulation. In this paper, the effectiveness of the proposed model was verified by a simulation using an evaluation agent instead of a user.