A Method of Sensitivity Analysis of Complex Systems Based on BP Neural Network

The paper proposes a method based on BP neural network to extract sensitivity information for the complex system with interrelated inputs. This method requires firstly establishing BP neural network models for each input to the others and system inputs to output, then sensitivity information is extracted from these BP network models. A procedure to obtain BP network models of the system with interrelated inputs is proposed; the question of whether for a certain sub-model to exist is explored; and a concrete approach to calculate sensitivity is presented. Simulation result demonstrates the validity of the proposed method of extracting sensitivity information.