Scholarlink

<p>Despite its challenges, the building sector remains a crucial pillar of Iraq's economic progress. This study aims to develop and evaluate machine learning models for predicting the productivity of housing projects in Iraq based on project design and implementation characteristics. Three advanced prediction techniques were used, including artificial neural networks (ANN), extreme gradient boosting (XGBoost), and support vector regression (SVR). Actual data from 67 housing projects were used, including seven input variables and one output variable represented by implementation productivity (real time). The evaluation results showed that the XGBoost model achieved the highest predictive accuracy with a determination coefficient of R<sup>2</sup> = 0.998 and a relative error (MAPE) of 0.8%. In comparison, the ANN model also performed well (R<sup>2</sup> = 0.92 and MAPE = 4.1%). In contrast, the SVR model performed less well with an accuracy of R<sup>2</sup> = 0.813 and MAPE = 6.6%. The results of the feature importance analysis also revealed that the most influential factors in productivity were built-up area, followed by the quantities of brick and concrete works. It can be implied that machine learning techniques, particularly XGBoost and ANN, provide effective and accurate tools to support decision-making in the early stages of housing project planning by predicting executive performance based on project data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Application of XGBoost and artificial neural networks in predicting housing project productivity

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月25日

查看全文
  • Ali Hassoon,
  • Farid Ezanee Mohamed Ghazali,
  • Tareq A. Khaleel

摘要

Despite its challenges, the building sector remains a crucial pillar of Iraq's economic progress. This study aims to develop and evaluate machine learning models for predicting the productivity of housing projects in Iraq based on project design and implementation characteristics. Three advanced prediction techniques were used, including artificial neural networks (ANN), extreme gradient boosting (XGBoost), and support vector regression (SVR). Actual data from 67 housing projects were used, including seven input variables and one output variable represented by implementation productivity (real time). The evaluation results showed that the XGBoost model achieved the highest predictive accuracy with a determination coefficient of R2 = 0.998 and a relative error (MAPE) of 0.8%. In comparison, the ANN model also performed well (R2 = 0.92 and MAPE = 4.1%). In contrast, the SVR model performed less well with an accuracy of R2 = 0.813 and MAPE = 6.6%. The results of the feature importance analysis also revealed that the most influential factors in productivity were built-up area, followed by the quantities of brick and concrete works. It can be implied that machine learning techniques, particularly XGBoost and ANN, provide effective and accurate tools to support decision-making in the early stages of housing project planning by predicting executive performance based on project data.