Assessing cost overrun in the saudi arabian building construction projects for value preservation

The pervasive issue of cost overruns remains a significant barrier to ensuring project success within budget and schedule constraints. While the causes of overruns are known, a major gap exists in providing validated, high-accuracy predictive tools for proactive expenditure control. This research addresses this gap by aiming to identify the most significant cost overrun factors in building construction projects and develop an accurate predictive model using advanced machine learning. The methodology involved collecting data via questionnaire surveys and interviews, assessing 49 pre-identified factors from industry experts. The Relative Importance Weight (RIW) method was used to prioritize these causes. Key findings revealed that 70% of the most significant factors were classified as internal project control deficiencies, including changes in scope and design, inadequate documentation, and poor site management. To achieve proactive forecasting, the study developed and systematically compared seven diverse Machine Learning (ML) architectures. The General Regression Neural Network (GRNN) was identified as the optimal model, achieving superior predictive performance with an average error (Mean Absolute Percentage Error, MAPE) of less than 3% and a correlation (R) of 0.95. The novel contribution of this work is the validated GRNN model, which transforms problem recognition into a reliable, actionable decision-support system for expenditure optimization and project value preservation in the construction sector.