Skip to main content

Table 1 General group of methods for data stream clustering

From: Data stream clustering by divide and conquer approach based on vector model

Method Pros Cons
Condensation-based [5, 2629] Having summary of data (global view)
Linear complexity
Scalability
Additive and subtractive property
Resource constraints
Detecting only spherical shape
Relearning
Applicable in low dimension
Data sampling [4, 30] Speed up
Memory usage
Low computational complexity
Low quality
Density –based [9, 15, 3133] Arbitrary shaped clusters Density threshold must be determined
Noise sensitivity
Outlier sensitivity
Applicable in low dimension
Relearning
Grid-based [2, 19, 34] Arbitrary shaped clusters
High dimension [14]
Stability
Relearning
Hierarchical structure [11, 19, 35, 36] Support evolving and concept drift
No need to determining extra parameters
Relearning
Inflexibility