改进的加权极限学习机不平衡分类算法及其应用OA

In blast furnace condition classification tasks,traditional classification models often suffer from low accuracy and poor robustness due to limited sample sizes and imbalanced class distributions.To address these challenges,this paper proposes an improved weighted extreme learning machine(WELM)algorithm for imbalanced classification.The method employs a Tent chaotic map-based non-dominated sorting genetic algorithm(CNSGA-Ⅱ)to optimize model parameters.Three key innovations are introduced:Firstly,the Tent chaotic mapping replaces random initialization to enhance the spatial uniformity of initial populations.Secondly,a dynamic boundary constraint mechanism is incorporated during genetic op-erations to suppress boundary violations while accelerating convergence.Thirdly,a dual-objective optimization framework simultaneously maximizes precision and recall rates,thereby improving minority-class recognition without compromising majority-class performance.Experimental validation using real-world blast furnace operation data demonstrates the effec-tiveness of the proposed method in handling imbalanced classification tasks.The results show significant improvements in both classification accuracy and model stability compared to conventional approaches.