the following statement reached me today - a generic topic I'd like to share with you:
Well, there's a significant difference between the airline and the routing distances. I also created some samples based on same input locations - just the DistanceMode is changed. Both samples have in commonI have a confusion – how beneficial is to use dima in planCluster service? As per my understanding, this service creates territories and just for creating territories how dima is useful? Suppose I don’t want to use estimatedTravelTimeToCluster, then does it make sense to use Dima with planCluster? Because dima is to use calculate driving time and driving distance.
- 300 input locations (heterogeneous activities)
- to be clustered into 5 clusters
- xCluster2.planClusters
If you apply airline you might get some "rough and obvious" structures but the quality is as good as within a sequencing/tour optimization based on airline: you might be able to identify obvious issues on an airline based model (such as loops and overlaps in a tour). But from a quality perspective only the routed distance matrix based approaches consider geographic aspects such as rivers, valleys, highways (or missing highways).Bernd
- 300 input locations
- The centers determined via airline / directDistance
- The centers determined via ExistingDistanceMatrix. The choice is different compared to the DirectDistance output
- The assigned locations based on DirectDistance
- The assigned locations based on ExistingDistanceMatrix
- From a pure perspective of the KPIs returned by xCluster the results seem to be equal.
- Here are the KPIs of the distance matrix itself (successful routing, average distance = 123.386km, average traveltime = 1:56:44)
- Now the DIMA based on airline. Average distance = 120.889km with average traveltime = 2:00:53
