Faith Dowelani has graduated with a doctorate specialising in Quantity Surveying from the University of Pretoria’s (UP) Faculty of Engineering, Built Environment and Information Technology (EBIT), combining her expertise in quantity surveying with machine learning to address one of the construction industry’s persistent challenges: inaccurate cost estimates at the early stages of building projects.
Dowelani’s PhD, completed in UP’s Department of Construction Economics, placed her among the University’s latest cohort of doctoral graduates who received their qualifications during the spring graduations, held from 31 August to 4 September 2026 on Hatfield Campus. Her research investigated whether supervised machine learning could improve the accuracy of cost estimates when limited project information was available.
Her study, titled ‘Supervised machine learning prediction models for early-stage cost estimation of building projects’, used historical data from 68 retail and office projects completed in Gauteng between 2013 and 2024. She developed and tested three prediction models to determine which could most accurately forecast the total capital investment required for a building project.
“At the conceptual stage of a project, detailed building drawings and specifications do not exist yet,” Dowelani explained. “Quantity surveyors have to produce cost estimates based on limited information, which is incredibly difficult and often leads to inaccurate estimates.”
Her models examined five key variables: the cost of the land, estimated construction or improvement costs, general development costs, financing costs and the type of building being constructed. She compared these supervised machine learning techniques: multiple linear regression, decision trees and artificial neural networks.
Dowelani found that machine learning could produce highly accurate early-stage cost estimates, even using a relatively small historical dataset. Multiple linear regression performed best, achieving a mean absolute percentage error of 3.52%, followed by the artificial neural network at 4.74%. The decision tree recorded an error of 22.64%.
One finding that surprised Dowelani was that the simpler multiple linear regression model outperformed the more complex artificial neural network.
“There is often an assumption that more advanced or complicated artificial intelligence models would automatically provide the best results,” she said. “My research showed that for this type of limited, context-specific data, a simpler model can be more accurate, practical and easier for professionals to interpret and trust.”
Dowelani said the findings showed that quantity surveying firms could begin using historical project data more systematically to improve future estimates. The models could be used as decision-support tools during the conceptual stage of a project, allowing quantity surveyors to generate an initial prediction of total capital costs before applying their professional judgement to project-specific factors.
The research journey was not without challenges. Dowelani said collecting building cost data was particularly difficult because such information is commercially sensitive. She also had to develop machine-learning skills despite coming to the PhD with a background in quantity surveying rather than computer science.
“Learning [computer programming software] Python, understanding the algorithms and modelling, and figuring out how to apply them to construction cost data required a steep learning curve,” she said.
Managing her PhD alongside other responsibilities was another challenge. Dowelani said the support of her family helped her persevere, particularly after the death of her father in 2023.
“My mother and sisters believed in me even when I doubted myself,” she said. “Their love and encouragement gave me the strength to keep pushing forward, even during the difficult moments.”
She describes the achievement as the realisation of a dream she had held since childhood, adding that her parents instilled in her the belief that education was the key to a better future.
“This achievement is not just mine as the first black South African to earn a PhD in quantity surveying at UP,” Dowelani said. “It also belongs to my mother, Alufheli, who sacrificed so much to support my education. She was the first woman to obtain a diploma in Architecture from the former Technikon Northern Transvaal, now Tshwane University of Technology; her pioneering spirit has been my greatest inspiration.”
Professionally, the PhD has positioned Dowelani to contribute to the future of quantity surveying through research, teaching, mentorship and the practical application of digital technologies in construction. Her ambition is to become the first female professor in quantity surveying at UP. She plans to continue her research, develop new models and approaches, train the next generation of quantity surveyors, work with industry partners to implement the models in practice, and contribute to discussions about technology in construction cost management.
“I hope that my journey inspires other young black women to know that they belong in academic spaces, research, leadership and professions where they may not always see themselves represented,” she said.
©Higher Education Media Services.