Predicting manpower allocation for landscape construction projects using XGBoost and SHAP

Purpose This study focuses on predicting manpower allocation for landscape construction projects using the eXtreme Gradient Boosting (XGBoost) algorithm combined with SHapley Additive Explanations (SHAP) to enhance workforce planning in response to growing labor shortages in Taiwan's construction industry. Design/methodology/approach Based on the convenient sampling concept, a dataset of 1,557 records was collected from public landscape projects executed in the North Taiwan, featuring major variables including project count, type, and calendar month. To enhance workforce planning in response to growing labor shortages in Taiwan's construction industry, the framework employs an interpretable labor forecasting model based on the proposed methods to improve workforce deployment efficiency and reduce the risks and cost overruns associated with labor shortages. Findings The results yielded by XGBoost and SHAP show that the proposed model achieved a high predictive accuracy (R2 = 0.892), and that labor demand is most influenced by project density, task type and seasonal variation. SHAP analysis further confirmed that the model provides interpretable insights into workforce trends. This forecasting framework offers practical value by enabling (1) early identification of peak labor periods for scheduling and outsourcing and (2) improves labor allocation with project demand in urban landscape planning. Originality/value These findings highlight the task-intensive and time-sensitive nature of landscape construction, where workforce scheduling often reflects project density as a core operational logic. Overall, the findings contribute to data-informed decision-making in construction workforce management.