A novel neural-network-related approach for regression analysis with interval model

We propose a new approach for interval regression using neural networks, which is different from two existing methods in the architecture of neural networks. Following the brief description of the existing neural network models and their learning algorithms for interval regression, we introduce a novel neural network model for interval regression that is a three-layer feedforward neural network with two output units, and then derive the corresponding learning algorithm. We finish some comparative experiments among three methods by means of a numerical example. Simulation results show that our approach with relatively simple network architecture can achieve approximate performance in comparison with other approaches. In addition, as an application we apply the proposed method to a real problem