Research on the Role of Feature Factors in Pneumoconiosis Staging Based on a Pneumoconiosis Database

China is the world’s largest coal producer, and pneumoconiosis remains the most widespread and severe occupational disease in the country. In recent years, artificial intelligence (AI) and radiomics technologies have attracted increasing attention in the diagnosis of coal workers’ pneumoconiosis (CWP). In this study, data extracted from the pneumoconiosis database were used to investigate the influence and contribution of multiple feature factors to disease staging through AI-based modeling. Using variables including age, dust exposure years, total working years, and smoking years, we extracted data from patients at different stages of pneumoconiosis. Two binary classification models were constructed using the XGBoost classifier: Model 1 (Stage I vs. Stages II–III, label: 0 vs. 1) and Model 2 (Stage II vs. Stage III, label: 0 vs. 1). The analysis revealed that in the early progression phase (Stage I → Stages II + III), Dust Exposure Years served as the principal driving factor, whereas in the later phase (Stage II → Stage III), Age became the most critical predictor. Total Working Years, although relatively weak in the early stage, demonstrated a moderate influence in later-stage progression, particularly contributing to the likelihood of Stage III development. Across both stages, Smoking Years functioned as a weakly synergistic factor, exerting only minimal influence on overall disease progression.