Leveraging AI for Construction Project Performance Predictions

Construction projects have traditionally faced poor performance, seldom measured using key performance indicators like those in earned value management, leading to frequent delays and cost overruns. The advent of Industry 4.0 and the digitization of the construction sector now provide stakeholders with vast amounts of data, which, paradoxically, overwhelms project managers who often resist modern tools such as artificial intelligence (AI) for data analysis. Given the rise of Industry 5.0, which emphasizes human-machine collaboration and smarter, more sustainable solutions, this paper aims to contextualize AI's role in this new paradigm. Various AI techniques are increasingly applied in project management to extract actionable insights and predict project performance. Artificial neural networks (ANNs), inspired by the structure and function of biological neural networks, are widely used due to their ability to model non-linear events. However, ANNs struggle with textual data, a common type in project management. To address this limitation, this research implements a novel approach using transformers, a key architecture in large language models like Generative Pre-Trained Transformers (GPT). Transformers can convert textual data, such as project descriptions, into vectors that ANNs can process, enabling more accurate predictions of project performance. This transformer-based method has been tested using open data on capital project schedules and budgets from the New York City School Construction Authority. Although still in the early stages of development, the approach has shown promise in reliably predicting project delays and cost overruns. This research demonstrates that integrating transformers with ANNs can enhance the ability of project managers to make data-driven decisions, potentially improving the overall performance and outcomes of construction projects.