Recently, I came across the book Data Science for Transport by Charles Fox, a lecturer at the University of Leeds UK. The book is for students of Transport Studies and for professionals working in the transportation industry who wish to apply Data Science skills to their work.
As I flipped through its pages, it was the contents of an endorsement for the book by, Dr Tom van Vuren, Divisional Director at Mott MacDonald, that struck a chord. In his endorsement, Dr van Vuren succinctly explains some of the issues ailing the transportation modelling techniques and practices and how they are ill-equipped to survive in a data-rich environment the world is moving towards.
He says, “Transport modelling practice was developed in a data-poor world, and many of our current techniques and skills are building on that sparsity. In a new data-rich world, the required tools are different, and the ethical questions around data and privacy are definitely different. I am not sure whether current professionals have these skills, and I am certainly not convinced that our current transport modelling tools will survive in a data-rich environment. This is an exciting time to be a data scientist in the transport field. We are trying to get to grips with the opportunities that big data sources offer, but at the same time, such data skills need to be fused with an understanding of transport, and of transport modelling. Those with these combined skills can be instrumental at providing better, faster, cheaper data for transport decision- making; and ultimately contribute to innovative, efficient, data-driven modelling techniques of the future”
Of course, Dr van Vuren, speaks in the context of the future of data science within the transportation industry, but in doing so, he also summarises some of the gaps we see both, in how data is acquired and perceived at the moment within the industry. There are 2.5 quintillion bytes of data created each day, and it is only increasing by the second. Today data is the world’s most reliable crystal ball, and in time we are going to see its use become ubiquitous in every walk of life. Transportation included.
At Intelematics, we work with many engineering firms, local councils and government departments, and provide them with the traffic speed, congestion and vehicle count data to help them solve real-world transport problems. Dr van Vuren’s comment is a validation of what we have understood about this sector and address through our products.
“Providing data that is equipped for data-rich environments”
Unfortunately, a significant section of the transport engineering community still relies on rudimentary means of gathering data such as manually counting cars or using Pneumatic tubes to evaluate traffic in the area. Such techniques are slow, cumbersome and expensive. Because of this, traffic evaluation generally lasts for a limited period (often only four or five days) in very specific areas which captures a small part of the story, or one tiny snapshot in time. With this limited data at hand, it is often hard to model a transport system of the future. How do you factor in the effect of seasons on traffic, school days and holidays, and other unseen disruptions?
Intelematics has been collecting and sharing real-time traffic information for over a decade, and we understand traffic and transportation. Our rich transport and traffic data is collected through hundreds of thousands of vehicles on the road and sensors located on smart infrastructure collected in real-time with collections as rapid as per second frequencies.
Having recorded over 5 billion points of data, we specialise in historical and real-time traffic data for traffic speed and congestion, traffic count (queues), and incident data. This holistic approach ensures that the most up-to-date traffic information is captured and recorded. At the same time, our comprehensive model enables us to offer historical traffic data at a granular level. By default, we provide data in 15-minute intervals. But this can be reduced to three-minute intervals if required.
Since accuracy is vital for ensuring reliable projections for transport projects, we ensure our data is detailed, complete and accurate.
We enrich our data using multiple proprietary sources and machine taught algorithms.
According to John Cardoso, Senior Product Manager, Intelematics, “What makes our traffic data better is that, while others model the data, we work on actuals. When we model our actual data, we do so using multiple dimensions: the road dimension, the vehicle perspective, the levels of congestion and even the demographics of the area (residential vs commercial). For example, when we do a modest AI training exercise, we input 200 million points of data and output 700 million. This is far beyond what most traditional traffic modelling techniques can take.
We have minimal thresholds of data quality that ensure our services are reliable. If one particular source of data doesn’t meet our minimum threshold, our system automatically calibrates the results to ensure quality and safeguards the data from being skewed.
You’ll never have to compromise on quality outputs whether you are using it for transport modelling, budgeting, forecasting, evaluation or planning.
Traditional methods of procuring road traffic data can be time-consuming and often require months of planning. Sometimes, unknown challenges can derail physical surveys or skew the data. The ongoing COVID-19 crisis is a great example – in the current scenario with movement restrictions and social distancing it can be challenging to conduct physical surveys and even if you manage to do it, the data might be skewed, considering road traffic right now is 49% lower than what it was last year around the same time.
Open data is the next best resource, but it’s not uncommon to find large gaps of time unaccounted for, or the data is a few years old. It doesn’t usually account for seasonality (it will give you an average taken across a full 12-month period), and when you get into it eighty per cent of your time is spent manipulating it into a useable format.
Intelematics has ensured transportation professionals have easy access to clean, accurate and historical data for more than 2000 Australian suburbs. It’s as recent as last month and arrives in their inbox in a matter of minutes. Not just that, the data files are available in a format that can be easily used in GIS and BI-based analytical tools for transport planning or for creating spatial regression models for transportation demand forecast. Rich road traffic data was never delivered this fast.
“Cheaper data for transport decision- making”
When it comes to data, cheap doesn’t always define the quality of the product.
To put things in perspective, it costs around $100,000 to conduct a physical traffic survey in a suburb to collect a week’s worth of traffic data. It costs much less to get richer data for a year or a month for a suburb online from the Intelematics traffic data store.
We want to ensure that transport-related decision making is relevant, quick and cost-effective for every traffic and transport professional out there, and our data is designed for any data-driven modelling techniques of the future used by them.
At Intelematics, we are excited to see that there is a shift towards data-driven, decision making within the transportation industry, and more and more traffic engineers are relying on data-driven modelling techniques.
Now more than ever, we need transportation planners and strategists who can look into the future and resolve the challenges of tomorrow.