BIM log mining for interactive target value design

Even with building information modeling (BIM) adoption, early-stage cost and schedule estimation requires a round-trip to external tools or specialists, decoupling assessment from design iteration. Such delayed assessment breaks the early feedback needed for target value design (TVD), limiting designers' ability to respond early to emerging cost and schedule overruns. This paper proposes an artificial intelligence (AI)-augmented BIM log mining framework that facilitates interactive TVD, focusing on total cost and construction duration. The proposed framework comprises three modules: (1) element-level state logging; (2) cost and duration estimation using shape-constrained machine learning regressors; and (3) in-authoring visualization of estimation trajectories. Validation on 93 projects shows mean absolute percentage error (MAPE) of 41.2% for cost and 21.3% for duration. The proposed t-dependent origin anchoring transformation addresses low-count extrapolation violations, yielding logically consistent estimates for early design states underrepresented during training. State-log-based feature updates keep interactive latency below 100 milliseconds, providing immediate feedback.