Classification Prediction of Operational Efficiency Based on Particle Swarm Algorithm Optimized Hybrid Kernel Extreme Learning Machine

Abstract

In this paper, a hybrid kernel extremum learning machine (PSO-HKELM) model optimized based on particle swarm algorithm is proposed for solving the classification prediction challenge of the synergistic optimization of environmental protection benefits and operational efficiency in industrial production. The experimental results show that the model significantly outperforms the traditional decision tree (0.733-0.799), random forest (0.817-0.955) and gradient boosting model (0.785-0.962) in the core metrics such as Accuracy (0.895), Recall (0.895), Precision (0.887) and F1 Score (0.876). 0.785-0.962), in which the classification accuracy is improved by 4.08% compared with the second best random forest. In particular, through the global parameter optimization of the particle swarm algorithm, the model achieves a high equilibrium F1 value of 0.876 between recall and precision (leading by 5.9%-13.3%), which effectively alleviates the contradiction of the hybrid kernel function's sensitivity to the sample category. Although the AUC value (0.926) is slightly lower than CatBoost (0.962), the model demonstrates superior generalization ability and robustness in combination with the systematic advantages of other metrics. The study shows that the synergistic mechanism of hybrid kernel architecture and population intelligence optimization strategy can significantly improve the classification performance of the extreme learning machine, which provides a solution to the multi-objective optimization problem in industrial scenarios that takes into account the model accuracy and stability, and is of practical guidance value for promoting the intelligent transformation and sustainable development of the manufacturing industry.