Clustering Flashcards
1
Q
Linkage methods
A
- Single linkage -> nearest neighbor
- Complete linkage -> furthest neighbor
- average linkage -> cluster centroids
- Ward‘s method -> middle point between relevant items
2
Q
Dendogram & Agglomeration schedule
A
- > „LENGTH“ of line -> power of cluster
- > aim at minimum amount of clusters with maximal power
- > inflection point in agglo schedule
- > biggest drop in coefficient value
3
Q
Two-stage clustering
A
- Hierarchical clustering to define number of clusters and initial centroids
- Non-hierarchical clustering with centroids as input