GridUnlock

Inspiration

Getting from one place to another can take 10 mins or 50 mins depending on the time of the day. We wanted to give insight into traffic at different times of the day and week to allow people to make informed decision on when and how they travel.

What it does

Our web application uses historical traffic data made available by the UK government with machine learning using TensorFlow and Databricks to forecast and highlight past and future traffic flow data throughout the day and year .

How we built it

We used a flask backend with TensorFlow and Databricks to process and learn from a dataset with more than one million rows and used GoogleMaps AI to display the traffic data on an interactive map.

Challenges we ran into

Finding quality data on historic traffic flow was very hard, it took us more than 5 hours to find a source that had sufficient coverage of the UK, coverage of time and we could use to produce traffic flow information. We were using several new technologies and experienced lots of problems with them.

Accomplishments that we're proud of

We were extremely ambitious and stepped out of our comfort zone, specifically in relation to the use of machine learning models and very complex frontend we apis.

What we learned

Planning and being methodical helps lots but sometimes it's possible to get stuck there.

What's next for GridUnlock

Find higher quality datasets and fine tune our machine learning model to more accurately predict traffic flow models, use other factors to predict traffic flow other than previous traffic flows, for example weather and events. Incorporate our predictions into a route planning algorithm or integrate them into google's route planning services.