A BOQ-Driven BIM-to-Python Workflow for Time-Dependent Construction Unit-Price Forecasting Under Inflationary Conditions

Cost estimation remains challenging under volatile market conditions and rapidly changing material prices. This study develops a BOQ-driven BIM-to-Python workflow for time-dependent construction unit-price forecasting by linking BIM-generated Bills of Quantities (BOQs) with a Python-based forecasting layer. The workflow is demonstrated using six structurally similar villas in Esentepe, Northern Cyprus, covering 33 construction-element price targets and historical price records from 1997 to 2023. For the principal recent-market case, seven observations from 2020 to 2023 were used to predict an external March 2024 holdout. Ordinary Least Squares (OLS) regression was compared with Random Forest, support vector regression, quadratic regression, and a last-observed-price persistence benchmark. Random Forest achieved the lowest external-holdout mean absolute percentage error (MAPE) of 5.45%, compared with 6.37% for persistence and 9.38% for OLS. Expanding-window validation across 99 historical predictions yielded MAPEs of 20.23% for persistence and 20.98% for OLS, indicating performance variation across market periods. Training-window sensitivity showed that the 2020–2023 OLS window produced the lowest March 2024 error among the historical windows examined. The 2024–2029 outputs are therefore presented as long-horizon trend scenarios rather than externally validated forecasts. The findings support a data-linked, BOQ-driven workflow that combines BIM project information with a model-agnostic forecasting layer while explicitly acknowledging uncertainty in long-term extrapolation from sparse price data.