A method for industrial data anomaly detection based on MIFHO-BP

Anomaly detection has an important impact on the development of industry. How to detect anomalies based on industrial data is an important research hotspot. At present, most studies use neural networks for anomaly detection. However, there are some problems with the neural network method, such as slow convergence and easy to fall into local optimality. To address these issues, this paper proposes an industrial data anomaly detection method based on modified improved Fire Hawk optimization algorithm (MIFHO) and the Back Propagation Neural Network (BP). On the one hand, this method improves the basic Fire Hawk optimization algorithm (FHO) by using tent chaos mapping and nonlinear composite adaptive inertia weight stochastic choice strategy to improve the algorithm's optimization ability. On the other hand, this study combines MIFHO with the BP neural network to construct the MIFHO-BP anomaly detection model. We conducted several comparative experiments. Experimental results show that MIFHO-BP exhibits superior performance compared with the comparison models, confirming the superiority of MIFHO-BP in anomaly detection of industrial data.