| description abstract | Unexpected uncertainties often arise during construction project execution, impacting performance measures such as time, budget, scope, and quality. This study addresses these cost and schedule challenges by creating an automated risk management model that utilizes natural language processing (NLP) techniques. NLP techniques are powerful tools that can process and analyze natural language data, allowing us to uncover valuable insights from textual data. This method enables the extraction of meaningful information from bid, contract, and change order documentation. The bidirectional encoder representations from transformers (BERT) model, a widely recognized transformer-based model, transforms words and phrases into numerical representations. After that, cosine similarity is used to assess the similarity between new and old projects. All these techniques allow us to predict potential costs and schedule changes for upcoming projects based on data from past similar projects. The research question of this study is: How can historical bidding and change order documents be utilized to forecast uncertainties in project cost and schedule for new projects? To address this question, the authors proposed an approach using NLP, BERT, and cosine similarity to extract the relevant data from past similar projects to forecast the cost and schedule changes for upcoming new projects, thus providing proactive insights for project management. Using a case study of 113 projects, out of which 20% were set aside for testing, the model achieved an accuracy of 78.30% in forecasting cost changes and 75.0% in forecasting schedule changes, with an overall accuracy of more than 75% in predicting changes. This finding demonstrates the model’s efficacy in anticipating project uncertainties, thus significantly contributing to improved project management. This data-driven approach to managing uncertainties ultimately enhances overall project success and performance by allowing construction professionals to anticipate and address potential risks and variations proactively. | |
