Can FME be used to produce heat maps?

  • 21 January 2013
  • 4 replies

Can FME be used to produce heat maps?  Kind of like this:



I have about 20 million points and I would like to show where the hotspots are.

4 replies

Badge +2
I'd say so.



One way it could potentially be done is to:
  • Use the HullAccumulator to create rubber bands around clusters of points. 
  • Perhaps the Generaliser to smoothen out the clusters
  • Colour code the clusters.
  • Use the HullAccumulator to create larger clusters by using a larger tolerance
  • Repeat the smoothing and colouring
  • Rasterise the resultant coloured clusters.
Note that you can use the PointOnAreaOverlayer to find out how many points are within a cluster.



Or of course, you could try to create a DEM via the DEMGenerator. If you set it up right, I'm sure it would look spectacular!


Badge +21



Several posts and workspaces related to this online. And it can be done quite generic!:



Badge +3

fme has a R caller.

On the relevant forum (R) there are multiple examples to clustering techniques and heat map generation.

I idd adapted the surface builder method and use standard Gaussian calculation for hotspot generation. Picture shows surface building, generalizer and then rebuilding surface (to make it look nice)

Gradient building and gradient bar use same calculations, just different transformers.

Table created with same values etc. (table unreadable in pic ass actual map is large 300dpi A0)

Map was created using Mapnick Rasterizer (With a lot of sweat !) 26min runtime.

Trying the R caller recently to see how it fairs.

This thread is a decade old but hopefully someone will see this - is it possible to use FME to create a HeatMap for KML (Google Earth)?