A Review of Subsequence Time Series Clustering

Table 3

The summary of postproof period dimensions.

Article	Problem	Method	Algorithm	Goal	Extent
[63]	The problem of time series clustering from a single stream	Motif discovery	MDL-based clustering	Creating meaningful result	No
[32]	The problem of time series clustering from a single stream	All methods	𝐾-means, motif discovery, MDL-based discovery	Producing correct results	[24]
[73]	Discovery motif with arbitrary length	Pattern discovery	𝐾-best motif discovery (𝑘BMD)	Developing the main idea of best motif	[80]
[74]	Length of motifs in finding time series motifs	Pattern discovery	Grammar induction algorithm	Developing a motif visualization system based on grammar induction	[81–83]
[33]	Meaningless outcomes as outputs based on inputs	Pattern discovery	Selective sequence time series (SSTS)	Achieving meaningful results	[24]
[31]	Predefined constraints values	Pattern discovery	Motif discovery	Eliminate the problem of predefined constraint values such as width of subsequences, by utilizing motif discovery algorithm	[32, 33]
[71]	Extracting and classifying shapes from very noisy real world time series	Pattern discovery	Motif discovery, noise test	A new method for shape extraction from time series	[75]
[57]	The difficulty of scaling a search to large datasets	Pattern discovery	God’s algorithm (GOAL), embedded-based search method (EBSM)	Search and mine massive time series for the first time	No
[76]	Invalid subsequence time series clustering	Partitioning clustering	Phase shift weighted spherical 𝑘-mean (PS-WS𝐾M)	Clustering unsynchronized time series	[5]
[77]	Difficulty of scaling search to large datasets	Pattern discovery	God’s algorithm (GOAL)	Search and mine truly massive time series for the first time	No