Please enable JavaScript.
Coggle requires JavaScript to display documents.
Cluster A: Conventional Clustering Algorithms (1. Partition (Low time…
Cluster A: Conventional Clustering Algorithms
1. Partition
Low time complexity, high computing efficiency
Interactively update cluster center till convergence
PAM, CLARA, CLARANS, K-Means, K-Medoids
Convex data suitable, local optimal, outlier sensitive, cluster numbers required
2. Hierarchy
Suitable for arbitrary data, high scalability
Merging of neighboring clusters to single cluster or reverse approach
BIRCH, CURE, ROCK, Chameleon
High time complexity, cluster number dependent
3.Fuzzy Theory
Objects belonging relationship in interval [0, 1]
Probability of belonging, high clustering accuracy
FCM , FCS, MM
Low scalability, local optimal, initial parameter sensitive clustering, cluster numbers required
4. Distribution
Same distribution to same cluster
Belonging probability, high scalability to changed distribution & cluster numbers
DBCLASD, GMM
Parameters influence clustering, high time complexity
5. Density
High density data belong to same cluster
High clustering efficiency, suitable for arbitrary data
DBSCAN, OPTICS, Mean-shift
Low quality for uneven data, big size memory for big data, parameter sensitive clustering