Posted on 17th January 2024

Analysing UK Vehicle Data using AI

Own an old car and ever wondered how rare it is? Perhaps you're a car dealer and interested in the size of the market for particular types of vehicles? There are many reasons to want to explore the UK's Driver and Vehicle Licensing Agency (DVLA) data.

Numerous websites exist to supply this information and they all get it from the DVLA. Anyone can also access this data, for free, in the form of a CSV data file. But unless you know your pivot tables from your vlookups you might not want to delve into that directly, and even if you do you may want to automate that process.

Well recently I tried a tool called Julius.AI that I was so impressed with for this particular use case I decided to write a little about it here.

tbd

That's a graph of the 10 most rapidly diminshing cars in the last 5 years in the UK. That is assuming the DVLA data is accurate! But it's pretty neat how quickly you can get results with this system.

Here's how I got this result and how you can analyse this, or any other CSV data, in minutes:

  1. Sign up for an account at Julius.AI.

  2. Download the DVSA make and model data here: https://www.gov.uk/government/statistical-data-sets/vehicle-licensing-statistics-data-files

  3. Add the CSV file to Julius either via the menu on the left hand side or pressing 'Add files' and 'Upload' in the chat window.

tbd

  1. Ask it a question. To get the above example I asked:

"What are the 10 most rapidly diminshing cars in the last 5 years"

then

"Can you add into the table the year each car first appeared (had a count greater than zero)"

then I pressed the 'Visualize' button to get the graph / chart.

tbd

That's pretty impressive! Looks like the MK1 Ford Focus is vanishing from our roads fast.

I'd say this is one of the best "chat with your data" solutions out there currently. I can see people using it for meaningful data analysis in a wide range of settings. It's a great way to make sense of data and an (arguably better) alternative to 'dashboards'... With the caveat that this is only as good as the data you have.

Of course, all AI based responses should be verified for accuracy if they are to be presented as factual data. Why I think this system generally works well is that it is made from multiple building blocks, including but not limited to AI, that are thoughtfully arranged together to form a solution to a relatively specific problem. Just 'bolting' AI on to existing products or implementing a very generalised chat-bot is unlikely to satisfy users' needs.