For this next stage on the TechBritain website we wanted to re-examine how we define and determine 'clusters'.
Does a city make a cluster?
In the past, we started with a city, manually defined its boundaries, and considered anything inside that border part of the 'Manchester' etc cluster. The solution was great for an initial guide, but we didn't like the rigidity/assumption that clusters are defined by and constrained by their prescribed boundaries.
So, if not cities, what makes a cluster?
For our new system, we wanted to remove the assumption that a city makes a cluster. We consider a cluster to be a group of companies and startups working in close proximity to one another and so we started there and looked at creating a new way of finding clusters.
We sought to develop a methodology for clustering companies and through evaluating and testing existing cluster analysis techniques we settled on a specific density based algorithm that will look for areas of high density based on the distance between points. It is these areas of high density that we define as clusters and after some tuning we ended up with 52 clusters across the UK based upon our current data set. You can now view the list on our clusters page. We're looking forward to sharing more of our clustering methodology in a future blog posts.
We originally were going to give the clusters generic names to differentiate them from our old 'cities' list but we decided that, while cities don't create clusters, they are a useful way of naming the clusters. As a result, we used the city or town closest to the centre of the cluster in order to determine the cluster's name.
This is still a new system and, with more study and experience, we may look at some refinements over time or may continue to adjust the clustering and critical density factors to try to make the clustering engine be as true to the reality of the communities across the UK as we can.