Duration prediction models for construction projects: In terms of cost or physical characteristics?

Abstract

The purpose of this study is to investigate and to validate the ‘Bromilow’s Time-cost (BTC) model’ and the ‘Love et al. ’s Time-Floor (LTF) model’ for Turkish public building projects, to check their appropriateness and to compare their prediction performances on project duration. The LTF model formulates the project duration in terms of gross floor area & floor numbers while the BTC model formulated it in terms of cost. Despite the consideration of the different characteristics, both models were inspired from Australian building projects. The study also presents ‘best-fit models’ as a benchmark for the BTC and the LTF models in order to fill the gap in construction duration predictions. The research data were collected from a total of 530 Turkish public building projects which were completed in seven geographic regions by performing an archive search work. 80% of the total data (i.e. from 424 projects) were used to establish the models while the remaining 20% (106 projects) were used for validation. Eight project characteristics of; total construction area, building height, gross floor area, floor numbers, contract sum, actual cost, contract duration and the actual duration were collected. Since there was no central archiving conduction to obtain these characteristics, more than a thousand tender documents were examined, and for 530 documents these characteristics were available. Consequently, it was found that the BTC model is superior to the LTF model, which confirms that ‘cost’ is a more significant predictor of the ‘duration’ than ‘gross floor area’ & ‘floor numbers’. However, the proposed ‘best-fit model’ indicated that the ‘area’ and the ‘height’ are the main predictors of the ‘duration’ as well as the ‘cost’.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 32 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.