Monitoring travel patterns to inform transport decisions

The goal

To support the Cobalt Business Park in North Tyneside, the biggest park of its kind in the UK, in developing its digital offering. Around 14,000 people work at Cobalt, which carries out an annual travel survey of workers to help inform transport decisions in the area, such as how many buses or cycle racks may be needed.

Nebula Labs was asked to help them automate this survey so that it could be accurate, fast and more cost effective.

 

The results

By combining its specialist app capabilities with data science expertise from the National Innovation Centre for Data (NICD), Nebula Labs was able to develop an app that could track journeys and automatically work out whether a worker is traveling on foot or by bicycle, car, bus or Metro train.

If the app is adopted it could put an end to  the need for an expensive physical survey. During the project they:

• Gathered 700 data points from worker journeys
• Adapted an off-the-shelf model to analyse the data
• Harnessed machine learning to improve outcomes
• Identified journeys correctly to between 80% and 96% accurac

Collecting Data

Although Nebula Labs has a great deal of expertise in digital technologies, it needed to enhance those skills with additional data knowledge to develop the survey app.

They initially approached the Arrow project at Newcastle University, which matches businesses seeking to innovate with academics and experts, and they were quickly matched with NICD.

Initially, Nebula Labs was advised to gather more data and so they recruited 28 volunteers on the park who downloaded the app so that journeys could be monitored over a three-month period. Although the data was anonymous, the app followed individual journeys, recording information such as longitude, latitude and speed, and provided around 700 data points for the project.

 

Analysing Data

The NICD team then worked with Nebula Labs to identify an off-the-shelf model that could be used to analyse the data and recommended how it could be adapted and improved with some additional coding.

McKee explains: “The machine learning aspects of the model can increase its accuracy. As you feed more data through it the machine can learn the patterns and look for more patterns. And the more patterns it identifies the more accurate it becomes.”

The model, for example, recognises that someone traveling 60mph is unlikely to be a cyclist or a pedestrian. But it also looks for more subtle patterns, for example people who are travelling faster but with stationary periods may be waiting on a platform.

By identifying such patterns in the data, Nebula Labs has been able to identify how staff are travelling to work with a great deal of accuracy and this has been achieved by monitoring only a small number of journeys. The company intends to carry out more work to increase this success rate even further by allowing the model to analyse a larger number of journeys.

Data Solution

McKee says: “The project went really well because NICD are obviously specialists at what they do and we are specialists at the technology that we work with. So, the two of us coming together was a really good combination and collaboration because we brought our skills from the app sector and they added their expertise in AI, data analysis and machine learning.”

The project has not only provided a possible solution to improve the Cobalt travel survey but has also provided Nebula Labs with a potential new product and they are considering how it might be used for other business parks and transport authorities.

 

---

"Nebula Labs possesses extensive technical knowledge, and is well placed to apply any algorithm or technical direction to their clients digital products."