Rubber bush is used in dynamic vibration absorber as dissipating devices in damping boring bar. These devices actually have to support radial load in compression when chattering occurs. Mastering the behavior of the radial stiffness of the rubber bush implies an accurate understanding of dynamic vibration absorber. The behavior is, however, complex due to the changeable cross-sectional shape and boundary conditions of the rubber bush. By using artificial neural network, the radial stiffness can be predicted efficiently. According to the authors' knowledge, simulations and tests on radial stiffness of the rubber bush under combined different cross-sectional shape and boundary conditions by using artificial neural network have not been performed yet. The purpose of this study is thus to find the law of radial stiffness of rubber bush under different cross-section shapes and axial precompression conditions. In order to achieve this aim, simulations and tests under different chamfering sizes and axial pre-compression by using artificial neural network were first carried out.