热镀锌钢卷力学性能影响因素之间关系复杂，限制了模型精度的提升。采用k-means算法利用化学成分属性对镀锌钢卷数据集进行聚类，将数据聚成三种模式簇实现样本的优选。利用梯度提升树算法，开展各模式数据集与不划分模式的全数据集下的力学性能建模研究，最后结合网格搜索与交叉验证方法进行模型参数优化。研究结果表明，分模式下模型MAE误差相比于全数据集建模平均减小0。85 MPa。参数优化后，各模式下 MAE 误差平均减少 5。19 MPa，RMSE 误差平均减少3。63 MPa，提高了预测模型精度。
Prediction of mechanical properties for hot dip galvanized steel coil based on clustering and GBDT
The relationships among the factors affecting the mechanical properties of hot-dip galvanized steel coils are complicated,which limits the improvement of the model accuracy.In this paper,the k-means algorithm is used to cluster the galvanized steel coil data set by using the chemical composition attributes,and the data set is clustered into three pattern clusters to filter samples.The gradient boosting tree algorithm is used to research on the mechanical performance modeling of each pattern data set and the full data set without pattern division.Finally,the model parameters are optimized by combining grid search and cross-validation methods.The results show that the average MAE error of the model in the sub patterns is reduced by 0.85 MPa compared to the full data set modeling.After the parameters are optimized,the average MAE error in each mode is reduced by 5.19 MPa,and the average RMSE error is reduced by 3.63 MPa,which improves the accuracy of the prediction model.
万方数据
维普
