基于MIC-RF随机森林的热轧带钢力学性能预测OA

To meet the increasing requirements for prediction accuracy of mechanical properties of hot-rolled strip steea prediction model for the mechanical properties of hot-rolled strip steel was constructed based on the the MIC-RF method,with yield strength(Re)and tensile strength(Rm)as the research objects.The Pauta criterion was used to preprocess the actual hot-rolling data and remove outliers.The maximum mutual information coefficient method(MIC)was employed to screen important influencing features as input variables for the model.The random forest algorithm was used to train the model and was compared with a random forest model without feature selection.It is shown that the prediction errors for Re and Rm based on the MIC-RF are both less than 1%,with coefficients of determination above 0.99,and tthe prediction accuracy is superior to that of the random forest model without feature selection.The proposed model has higher prediction accuracy and strong generalization ability,providing an intuitive and reliable reference for the design and optimization of strip steel products.