Abstract: To improve the accuracy of the predicted width of the first piece of steel after changing the steel type， the steel specification and the roll in the process of hot continuous rough rolling strip production， a new width prediction model based on the principal component analysis collaborated with random forest (PCA-RF) algorithm is proposed in this work. The PCA method is used to analyze the reasonability of data samples and the feature selection is carried out by calculating the eigenvalue， and principal component and cumulative contribution degrees. The best RF model is trained on variant dataset selected from the PCA. At the same time， support vector machine regression(SVR)and K-nearest neighbor(KNN)models are used for comparison and verification. The practical applications show that the R-squared value from the each pass width predicted by the PCA-RF model is controlled within the range of 0.999~1， and the prediction error of more than 96% samples is -5~5mm， which proves that the model can predict the steel width with a high precision.

Key words: hot continuous rough rolling; principal component analysis(PCA); feature selection; width prediction; random forest(RF) algorithm