Project managers need reliable predictive analytics tools to make effective project intervention decisions throughout the project life cycle. This study uses Machine learning (ML) to enhance the reliability in project cost forecasting. A XGBoost forecasting model is developed and computational experiments are conducted using real data of 110 projects representing 1268 cost data points. The developed model performs better than some Earned value management (EVM), ML (Random forest, Support vector regression, LightGBM, and CatBoost), and non-linear growth (Gompertz and Logistic) models. The model produces more accurate estimates at the early, middle, and late stages of the project execution, allowing for early warning signals for more effective cost control. In addition, it shows more accurate estimates in most projects tested, suggesting consistency when repeatedly used in practice. Project forecasting studies mainly used ML to estimate the project duration; a few ML studies estimated the project cost at the project's conceptual stage. This study uses real data and EVM metrics, proposing an effective XGBoost model for forecasting the cost throughout the project life cycle.

A machine learning study to improve the reliability of project cost estimates / Narbaev, Timur; Hazir, Öncü; Khamitova, Balzhan; Talgat, Sayazhan. - In: INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH. - ISSN 0020-7543. - 62:12(2024), pp. 4372-4388. [10.1080/00207543.2023.2262051]

A machine learning study to improve the reliability of project cost estimates

Timur Narbaev;
2024

Abstract

Project managers need reliable predictive analytics tools to make effective project intervention decisions throughout the project life cycle. This study uses Machine learning (ML) to enhance the reliability in project cost forecasting. A XGBoost forecasting model is developed and computational experiments are conducted using real data of 110 projects representing 1268 cost data points. The developed model performs better than some Earned value management (EVM), ML (Random forest, Support vector regression, LightGBM, and CatBoost), and non-linear growth (Gompertz and Logistic) models. The model produces more accurate estimates at the early, middle, and late stages of the project execution, allowing for early warning signals for more effective cost control. In addition, it shows more accurate estimates in most projects tested, suggesting consistency when repeatedly used in practice. Project forecasting studies mainly used ML to estimate the project duration; a few ML studies estimated the project cost at the project's conceptual stage. This study uses real data and EVM metrics, proposing an effective XGBoost model for forecasting the cost throughout the project life cycle.
File in questo prodotto:
File Dimensione Formato  
A machine learning study to improve the reliability of project cost estimates.pdf

accesso aperto

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Creative commons
Dimensione 2.84 MB
Formato Adobe PDF
2.84 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2989064