Method for fusion of neighborhood rough set and XGBoost in welding process decision-making

Correct decision-making rules are essential to achieve the application of knowledge. The welding procedure document requires a rigorous knowledge rule system. Due to the limitations in representing and extracting practical engineering knowledge, the construction of knowledge rules is complicated. This paper proposed a synergistic approach of fusion model and interpretation analysis. The fused model uses neighborhood rough sets and XGBoost to refine knowledge and constructs implicit relationships. Common logic rules and knowledge are replaced with the model. The model was validated and analyzed based on standardized high-speed train bogie framing engineering data, and the scores obtained were 0.89 for accuracy, 0.92 for Precision, 0.89 for Recall, and 0.89 for F1-score. Based on ensuring the metrics of the model, the interpretable analysis method expresses the implicit knowledge in the decision-making system. The tree model is used to explain the decision process, and the relationships of the attributes involved in the decision can be obtained via SHAP analysis. Moreover, it shows a high degree of consistency between interpretable results and actual engineering knowledge. The experimental results indicate that the proposed method can be effective for intelligent decision-making in welding procedure documentation.