Random forest and path diagram taxonomies of risks influencing higher education construction projects

Abstract

While risk factors are sine qua non for construction projects’ non-performance, the research efforts are directed toward the likelihood of risks at the detriment of their level of influence on higher education building projects. This study assessed the perceptions of construction key stakeholders about the influence of risk factors on higher education building projects using machine learning-based random forest classification. A questionnaire survey was administered to four hundred and sixty-five (465) respondents comprising clients’ representatives, consultants, and contractors across five (5) higher education institutions in Nigeria. Of 465 questionnaires, 295 retrieved were suitable for the analysis implying a 63.44% response rate. The Random Forest (RF) classification used 295 samples, out of which 189 (64%) formed the training dataset, while the validation and testing data sets are 47 (16%) and 59 (20%), respectively. The RF model accuracy conducted shows the optimized model with the test accuracy and out-of-bag accuracy (OOB). The study clustered 58 risk factors into four comprising (i) security, access, health, and safety risks, (ii) construction dispute resolution risks, (iii) construction planning and contract documentation risks, and (iv) construction cost and management risks. Further, the proposed recommendations could help enhance the performance of higher education building projects.

KEYWORDS:

• Construction projects
• Nigeria
• random forest classification
• performance
• risk factors

Disclosure statement

No potential conflict of interest was reported by the authors.