基于改进BO-KNN-SVM的特种纸质量预测OA

To address the problems of strong nonlinearity,multi-parameter coupling,and insufficient generalization ability of single models in specialty paper quality prediction,this paper aims to construct a high-precision quality prediction model.A K-nearest neighbors(KNN)-SVM combined quality prediction model(TCBO-KNN-SVM)based on Tent chaotic initialization and Cauchy mutation improved Bayesian optimization(BO)is proposed.First,through feature selection,10-dimensional key process parameters were extracted from specialty paper production data,and data augmentation was combined to improve sample quality.Second,Tent chaotic mapping was used to optimize the uniformity of BO's initial sampling,and Cauchy mutation was introduced to enhance the global optimization capability of BO in the later iteration stage,thereby constructing the improved Tent chaotic initialization and Cauchy mutation-based Bayesian optimization(TCBO).Finally,the advantages of KNN's local fitting and SVM's global mapping were integrated,and the hyperparameters of the combined model were optimized via TCBO to realize specialty paper quality prediction.Experiments were conducted based on the actual production data of a certain enterprise.The results show that the TCBO-KNN-SVM model achieves coefficients of determination(R2)of 0.9782 and 0.9769 for tensile strength and air permeability prediction,respectively.Compared with the benchmark models(BO-KNN-SVM,PSO-KNN-SVM,BO-KNN,and BO-SVM)and PSO,the R2 of the proposed model is increased by an average of 2.00％—6.72％,while the root mean square error(RMSE)and mean absolute percentage error(MAPE)are both reduced by more than 20％.This model effectively improves the accuracy and stability of specialty paper quality prediction and can provide technical support for production quality control.