Cost data for transport infrastructure projects is recorded in a myriad of different ways. It varies from country to country, project to project and even within the same organisation.
If we could better compare cost data, we could compare projects, learn lessons from different approaches, and ultimately improve the way we deliver future projects and the value returned on investments. This research aimed to investigate whether artificial intelligence (AI) could be deployed to search through files of project cost data, including historical data, to find the relevant information and then process and classify it. Classification would be in line with the Royal Institution of Chartered Surveyors’ (RICS’) International Cost Management Standard (ICMS).
The researchers used information from 54 National Highways projects, 35 major motorway projects, the others renewal and maintenance works. This resulted in over 50,000 pieces of data. One of the benefits of using data from National Highways is that the infrastructure owner has already developed a conversion route from its internal system to the ICMS framework.
The AI tool developed by the researchers deployed a combination of natural language processing (NLP) and machine learning algorithms to sort through text, convert it to data and then match it up to categories from the ICMS. Human reviewers then checked the classification, with their feedback helping to inform the machine learning.
There are still improvements which could be made. For example, teaching AI how to deal with typographical errors and a multitude of abbreviations. However, the results were good. The researchers report an accuracy of between 80 and 95% for the AI tool when tested on the National Highways data, which they say is step-change compared to traditional analysis methods. Processing time was cut by up to 90%.