A Hybrid Algorithm for Clustering of Time Series Data Based on Affinity Search Technique

Table 1

Quality of TTC approach against the standard 𝑘-Medoids with regard to raw time series data and the time series data represented by PAA.

Dataset	Number of classes	DS size	Length	𝑘-Medoids (ED)		𝑘-Medoids (PAA-ED)		TTC
				RI	ConEnrtropy	RI	ConEnrtropy	RI	ConEnrtropy
50words	50	455	270	0.95	0.73	0.95	0.73	0.96	0.79
Adiac	37	391	176	0.94	0.58	0.94	0.58	0.96	0.64
CBF	3	900	128	0.67	0.29	0.7	0.41	0.88	0.78
Coffee	2	28	286	0.8	0.6	0.78	0.54	0.86	0.68
ECG200	2	100	96	0.61	0.18	0.61	0.18	0.55	0.14
FaceFour	4	88	350	0.77	0.51	0.77	0.51	0.91	0.78
FISH	7	175	463	0.77	0.32	0.8	0.32	0.83	0.48
SwedishLeaf	15	625	128	0.9	0.54	0.9	0.54	0.9	0.58
synthetic_control	6	300	60	0.82	0.56	0.87	0.79	0.92	0.95
Trace	4	100	275	0.75	0.51	0.78	0.59	0.85	0.78
Two_Patterns	4	4000	128	0.63	0.03	0.69	0.32	0.86	0.88
Wafer	2	6164	152	0.43	0.01	0.43	0.01	0.5	0.21