Research on predicting the productivity of cutter suction dredgers based on data mining with model stacked generalization

To solve the problem that dredging prediction systems provide inaccurate productivity predictions and rely heavily on mud concentration data. This paper presents a data mining method to accurately predict dredger productivity by using model-stacked generalization in the absence of mud concentration data. First, eliminate abnormal construction data, and l(2) norm normalization and log smooth transformation are then performed on the data. Second, Spearman's rank correlation coefficient method is used to extract features. Five machine learning models, namely, Lasso, Elastic net (ENet), Gradient-boosting decision tree (GBDT), extreme gradient boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM), were introduced to predict dredger productivity. Based on these five models, a stacked generalization model was applied. The results show that the goodness of fit R-2 of the stacked generalization model for productivity prediction is 0.9281, which is higher than the accuracy of the other algorithms investigated, and the optimization effect is obvious.