Cluster C: Recent Clustering Algorithms
6. Affinity Propagation
AP - potential cluster centers, Affinity - negative Euclidean distance, greedy strategy maximizes clustering global function value per iteration
AP clustering
Simple algorithm, insensitive to outliers, cluster numbers required
High time complexity, unsuitable to large data, parameters sensitive clustering
7. Density and Distance
DD clustering
High local density, away points with high local density
Simple algorithm, suitable to arbitrary data, outliers insensitive
High time complexity, cluster center based on decision graph, parameters sensitive clustering
8. Spatial Data
Large scale characteristics sharing, high speed, information complexity
DBSCAN, STING, Wavecluster, CLARANS
9. for Data Stream
Large scale sequence based characteristics sharing with limited reading
STREAM, CluStream, HPStream, DenStream
10. for Large-Scale Data
Considers 4 V's of data characteristics
K-means, BIRCH, CLARA, CURE, DBSCAN, DENCLUE, Wavecluster & FC