GWO-BP-AdaBoost: Enhancing Classification Performance of BP Neural Networks Through GWO-Based Optimization and AdaBoost Integration
Enqi Cao, Fanliang Bu · 2023
This paper proposes a method based on the Grey Wolf Optimizer (GWO) to automatically search and quickly determine the number of hidden nodes in a single-hidden-layer Back Propagation (BP) neural network, as well as the number of training iterations. This method aims to address the difficulties and suboptimal selection of the hidden layer node count and training iterations in BP neural network, which often lead to unstable predictive performance and unsatisfactory classification results. By optimizing the structure and training process of the BP neural network, the proposed approach enhances its classification capability. In addition, by employing the Adaptive Boosting (AdaBoost) algorithm to combine multiple weak classifiers, which are GWO-BP neural networks, a GWO-BP-AdaBoost strong classifier model is constructed to further enhance the classification performance of the BP neural network. Experimental results demonstrate that the classification ability of the BP neural network is enhanced through optimization using the GWO algorithm. Moreover, the subsequently established GWO-BP-AdaBoost classification model exhibits further improvement in classification ability compared to the GWO-BP neural network. The GWO-BP- AdaBoost classification model effectively classifies the dataset with high accuracy and excellent classification performance.