Neighbor Attribute Average

From fmepedia

The attached workspace uses FME to find the average value of an attribute for all neighboring polygons.

The Workspace

Above: Neighbor Attribute Average Workspace.

The workspace (created in FME2008) generates test data in the form of a set of polygon features with an attribute (set to random values). It finds the average value of the attribute for all neighboring polygons using the method:

Intersect polygons to create a line for each common segment. The line will have a list of component polygons.
Sum the list of neighbor values and the number of items in each list (1 for a non-adjacent polygon).
Explode the list back into its component parts
Deduct the original polygon's value from the sum (we only want to count neighbors)
Aggregate all boundaries back into the original polygon (using a feature ID) and creating a sum of all neighbor values
Calculate the average neighbor value (sum of all values / number of neighbors)

The attribute used can be any numeric value, for example a DEM Z value obtained with the CoordinateFetcher.

Above: The results. Your results would not be the same, since this is generated from random numbers.

Note that the process does not take account of diagonally adjacent polygons (single point match), only those with a common line segment (two point line or polyline).