First of all let me mention that a cluster planning tries to satisfy several goals:
- Geographic compactness of the cluster areas
- Balanced workload within the clusters (aka distribution)
The overall process of a cluster planning is splitted into different phases:
- ConstructionPhase – during this phase the focus is on compact areas. The workload balance is ignored during this phase. We recommend to execute this phase whenever you perform a new clustering.
- ImprovementPhase – during this phase the algorithm tries to improve a given solution by swapping visits between clusters. Compactness and Workload are goals during this phase.
- PostOpt – this special phase tries to improve the compactness again by swapping orders with equal “workload” between the clusters. Furthermore this only affects elements with a single visit during the planning cycle. This means that the distribution of the clusters doesn’t change (because of the “same workload” restriction)
- Distribution does not matter? improvementCompactnessFactor = 100 , improvementDistributionFactor = 0
- Compactness and distribution should be treated almost equal? improvementCompactnessFactor = 100 , improvementDistributionFactor = 100
- Prio A is compactness but also distribution should at least be taken into consideration as a Prio B goal:
improvementCompactnessFactor = 100 , improvementDistributionFactor = 10