Performance prediction of construction projects using soft computing methods

Seyedeh Sara Fanaei, Osama Moselhi, Sabah T. Alkass

Research output: Contribution to journalArticlepeer-review

6 Citations (Scopus)

Abstract

Key performance indicators (KPIs) evaluate different aspects of projects and are used to determine the health status of projects. While there is considerable work on project quantitative performance prediction, less attention, however, has been directed towards qualitative performance prediction. This paper offers a novel framework for qualitatively measuring and predicting six important construction project KPIs using the neuro-fuzzy technique. Neuro-fuzzy models are developed to map the KPIs of three critical project stages to whole project KPIs. Subtractive clustering is utilized to automatically generate initial fuzzy inference system (FIS) models and the artificial neural network (ANN) technique is used to tune the parameters of the initial FIS models. The relative weight of each KPI is determined using a series of computing methods namely, analytic hierarchy process (AHP) and genetic algorithm (GA), to generate the performance indicator (PI). The developed models are validated with real project data showing that the rate of error is reasonably low. The results show that the AHP method is more accurate when compared to the GA method. This framework can be used in building construction projects to help decision-makers evaluate the performance of their projects.

Original languageEnglish
Pages (from-to)609-620
Number of pages12
JournalCanadian Journal of Civil Engineering
Volume46
Issue number7
DOIs
Publication statusPublished - 2019
Externally publishedYes

Keywords

  • Construction project
  • Key performance indicators (KPIs)
  • Neuro-fuzzy
  • Performance forecasting

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Environmental Science(all)

Fingerprint

Dive into the research topics of 'Performance prediction of construction projects using soft computing methods'. Together they form a unique fingerprint.

Cite this