A Knowledge Transfer Partnership between the University of Nottingham and construction consultancy Gleeds is bringing the power of AI to streamline the process for cost estimation.

The Knowledge Transfer Partnership (KTP) is an innovative collaboration that connects academia and industry in our fast-paced world.

Funded by Innovate UK, the KTP aims to infuse industries with the knowledge, expertise and techniques developed in academic institutions.

This collaboration fosters an environment ripe for innovation, enabling businesses to drive economic growth more efficiently. A prime example of this successful partnership is the one between the University of Nottingham and Gleeds.

In the construction industry, client interactions and cost prediction are crucial. During the initial stage of any project, known as RIBA stage 0/1, clients present their project visions and we provide cost estimates — a process that continues until both parties agree on a budget.

Recognising the need for a more efficient approach, we sought an AI-driven solution capable of estimating costs instantly, allowing real-time adjustments during client meetings.

A reliable database was needed to harness the power of AI

However, our pursuit of this solution unveiled the need for a reliable database to power artificial intelligence. Existing construction datasets contained inconsistencies that could disrupt any machine-learning algorithm.

Instead of opting for quick fixes, we embarked on a journey with the University of Nottingham to collect and validate data meticulously, prioritising quality over quantity.

Selecting projects for data collection was critical. We found that larger construction projects provided rich, quality data, highlighting stark differences – both between and within sectors.

We likened this variability to distinguishing between different animal species and understanding the nuanced variations within them – akin to discerning what makes a residential building a ‘cat’ and a hotel a ‘dog’ – and then identifying the subtle differences within each category.

A KTP is a game-changer for data extraction

After building a robust, construction industry-specific database, we turned our attention to another challenging task: manually extracting data from cost files.

This process was not only tedious but also time-consuming, sometimes taking up to 16 hours per file.

I was fortunate to have an MSc student join me from the University of Nottingham and a game-changing solution emerged: a tool that could automatically extract data from more than 80% of the files, reducing extraction time to just 30 minutes to an hour.

This tool offered an added advantage. It was capable of automatically classifying item descriptions according to New Rules of Measurement (NRM) elements.

This hierarchical structure, widely used in the construction industry for cost estimation, allows us to build up costs from granular rates to broader categories, addressing the issue of “bulked together” costs that had previously hampered cost estimation.

Cost prediction using machine learning

Ultimately, armed with a meticulously crafted dataset and cutting-edge tools, we embarked on the pinnacle of our Knowledge Transfer Partnership journey: cost prediction using machine learning.

Despite the complexities and variations inherent in construction data, our machine learning model achieved a respectable prediction accuracy of up to 75%, with a less than 10% error rate. We consider this to be a starting point. As we gather more data, the model will improve over time.

Beyond technological solutions, the KTP journey has helped to transform Gleeds into a data-driven business, prompting the creation of a sophisticated benchmarking tool, a rates database and a data academy to upskill employees.

This transformation has also sparked a cultural change within our organisation, as employees are now empowered to devise their own solutions, igniting a wave of innovation throughout.

Is the KTP an example of what collaboration can achieve?

Our experience with the Knowledge Transfer Partnership exemplifies its transformative power. It highlights how the fusion of machine learning and data engineering can revolutionise traditional industries, enhancing efficiency and accuracy, driving innovation and nurturing a culture that values data and continuous learning. The KTP serves as a model for what collaborative efforts between academia and industry can achieve.

First published on pbc today on the 9th June 2023

Get in touch
 Alexandra Spencer

Alexandra Spencer
Data scientist