AI-driven Cost Estimation Accuracy in Construction Projects

Thesis Abstract

Effective cost estimation remains a critical challenge in construction projects due to the inherent uncertainties and complexities associated with project scope, resource allocation, and market fluctuations. Traditional estimation methods often result in cost overruns, project delays, and compromised financial planning, thereby emphasizing the need for innovative, data-driven approaches. Recent advancements in artificial intelligence (AI) technologies provide promising avenues to enhance the precision and reliability of cost estimates, yet empirical evidence on their efficacy within construction contexts remains limited. This study aims to evaluate the impact of AI-driven models on the accuracy of cost estimation in construction projects, with specific focus on identifying the key factors that influence predictive performance and establishing best practices for implementation. The primary objective of the research is to develop and validate an AI-based cost estimation framework that outperforms conventional methods. To achieve this, the study also seeks to compare the accuracy of AI models—including machine learning algorithms such as random forests, support vector machines, and neural networks—against traditional statistical approaches like linear regression. Additionally, the research explores the influence of project typology, complexity, and data quality on model performance. The study further aims to generate insights into the integration of AI systems within existing project management processes to facilitate decision-making and cost control. Methodologically, the research adopts a mixed-methods approach, combining quantitative and qualitative data collection and analysis. The quantitative phase involves a cross-sectional survey of 150 construction firms operating across the region, selected through stratified random sampling to ensure representativeness across project sizes and types. Data collection instruments include structured questionnaires designed to elicit factors influencing cost estimation practices, alongside case study data from 50 recent projects where AI applications were implemented. The case data encompass project cost records, resource inputs, and project outcome metrics, collected through interviews with project managers and review of project documentation. Quantitative data will be analyzed using multiple regression analysis and analysis of variance (ANOVA) to assess the predictive performance of AI models versus traditional methods, as well as to examine factors affecting estimation accuracy. Model validation will utilize metrics such as Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and R-squared values to evaluate predictive accuracy. Qualitative data from interviews will be analyzed thematically to extract insights on challenges, adoption barriers, and perceptions regarding AI integration, guided by the Diffusion of Innovations theory and Technology Acceptance Model (TAM). The anticipated findings suggest that AI-driven cost estimation models can significantly improve predictive accuracy compared to conventional approaches, particularly when high-quality, comprehensive datasets are employed. The study expects to demonstrate that machine learning algorithms, especially neural networks, outperform traditional statistical models in capturing complex, non-linear relationships in project cost variables. Furthermore, factors such as project complexity and data integrity are hypothesized to influence the models’ effectiveness. The integration of AI systems into existing planning and control processes is anticipated to enhance decision-making efficiency, reduce risk, and foster greater confidence among project stakeholders. This research contributes to knowledge by providing empirical evidence of AI’s potential to revolutionize cost estimation practices in construction, offering a validated framework for model development and implementation. It also advances understanding of factors influencing AI adoption within project environments, guiding best practices for technology integration. In conclusion, the study recommends the adoption of AI-driven estimation models as standard practice, supported by robust data management systems and stakeholder training, to mitigate cost overruns and improve project delivery outcomes. Future research directions include exploring real-time AI applications and integrating cost estimation models with Building Information Modeling (BIM) platforms for enhanced project visualization and control.