Methodological Framework for Forecasting Directive Changes in Project Calendar Models

This article develops a methodological framework for forecasting directive changes in schedule models of investment and construction projects. Directive changes are understood as mandatory managerial interventions initiated by the client and implemented outside formalized change management procedures. Such interventions are considered as an independent object of analysis exerting a systemic influence on the stability and dynamics of project schedule-network models. The study is based on the integration of a directive change classifier and index-based characteristics of the directive flow with a system of consequence indicators intended for the quantitative assessment of the schedule effects caused by directive interventions. The XGBoost algorithm is proposed as the core forecasting instrument due to its ability to identify nonlinear relationships between directive parameters and the response of the project scheduling system. Interpretation of the forecasting results is supported through SHAP analysis, enabling a quantitative assessment of the contribution of each classifier feature to the generated forecast. The proposed methodological framework includes sequential stages of directive formalization, feature encoding, model training, interpretation of results, and scenario-based integration of forecasts into project schedule models. The article also substantiates the possibility of digital integration of forecasting results into PMIS, MS Project, and 4D BIM environments, thereby increasing the transparency of schedule models and expanding the analytical capabilities of scenario-based assessment of directive changes in investment and construction projects.