时间序列数据集：UCR Time Series Classification Archive【共128个数据集】

UCR是时间序列数据集，并且每个数据集样本都带有样本类别标签，目前是时间序列挖掘领域重要的开源数据集资源。

UCR Time Series Classification Archive数据集

在2018版的官网页面上可以直接下载整个128个数据集，下图中的红框1可以阅读PDF文档，红框2是下载按钮。

一、Data Format

Each of the data sets comes in two parts, a TRAIN partition and a TEST partition.
For example, for the Fungi data set we have two files, Fungi_TEST.tsv and Fungi_TRAIN.tsv
The two files will be in the same format but are generally of different sizes.
The files are in the standard ASCII format that can be read directly by most tools/languages.
For example, to read the data of Fungi data set into MATLAB, we can type…

>> TRAIN = load(‘Fungi_TRAIN.tsv'); 
>> TEST  = load(‘Fungi_TEST.tsv' ); 

…at the command line.

There is one time series exemplar per row. The first value in the row is the class label (an integer between 1 and the number of classes). The rest of the row are the data sample values. The order of time series exemplar carry no special meaning and is in most cases random. A small number of data sets have class label starting from 0 or -1 by legacy.

官网首页如下：

ID	Type	Name	Train	Test	Class	Length	ED (w=0)	DTW (learned_w)	DTW (w=100)	Default rate	Data donor/editor
1	Image	Adiac	390	391	37	176	0.3887	0.3913 (3)	0.3964	0.9591	A. Jalba
2	Image	ArrowHead	36	175	3	251	0.2000	0.2000 (0)	0.2971	0.6057	L. Ye & E. Keogh
3	Spectro	Beef	30	30	5	470	0.3333	0.3333 (0)	0.3667	0.8000	K. Kemsley & A. Bagnall
4	Image	BeetleFly	20	20	2	512	0.2500	0.3000 (7)	0.3000	0.5000	J. Hills & A. Bagnall
5	Image	BirdChicken	20	20	2	512	0.4500	0.3000 (6)	0.2500	0.5000	J. Hills & A. Bagnall
6	Sensor	Car	60	60	4	577	0.2667	0.2333 (1)	0.2667	0.6833	J. Gao
7	Simulated	CBF	30	900	3	128	0.1478	0.0044 (11)	0.0033	0.6644	N. Saito
8	Sensor	ChlorineConcentration	467	3840	3	166	0.3500	0.3500 (0)	0.3516	0.4674	L. Li & C. Faloutsos
9	Sensor	CinCECGTorso	40	1380	4	1639	0.1029	0.0696 (1)	0.3493	0.7464	physionet.org
10	Spectro	Coffee	28	28	2	286	0.0000	0.0000 (0)	0.0000	0.4643	K, Kemsley & A. Bagnall
11	Device	Computers	250	250	2	720	0.4240	0.3800 (12)	0.3000	0.5000	J. Lines & A. Bagnall
12	Motion	CricketX	390	390	12	300	0.4231	0.2282 (10)	0.2462	0.8974	A. Mueen & E. Keogh
13	Motion	CricketY	390	390	12	300	0.4333	0.2410 (17)	0.2564	0.9051	A. Mueen & E. Keogh
14	Motion	CricketZ	390	390	12	300	0.4128	0.2538 (5)	0.2462	0.8974	A. Mueen & E. Keogh
15	Image	DiatomSizeReduction	16	306	4	345	0.0654	0.0654 (0)	0.0327	0.6928	ADIAC project
16	Image	DistalPhalanxOutlineAgeGroup	400	139	3	80	0.3741	0.3741 (0)	0.2302	0.5324	L. Davis & A. Bagnall
17	Image	DistalPhalanxOutlineCorrect	600	276	2	80	0.2826	0.2754 (1)	0.2826	0.4167	L. Davis & A. Bagnall
18	Image	DistalPhalanxTW	400	139	6	80	0.3669	0.3669 (0)	0.4101	0.6978	L. Davis & A. Bagnall
19	Sensor	Earthquakes	322	139	2	512	0.2878	0.2734 (6)	0.2806	0.2518	A. Bagnall
20	ECG	ECG200	100	100	2	96	0.1200	0.1200 (0)	0.2300	0.3600	R. Olszewski
21	ECG	ECG5000	500	4500	5	140	0.0751	0.0749 (1)	0.0756	0.4162	Y. Chen & E. Keogh
22	ECG	ECGFiveDays	23	861	2	136	0.2033	0.2033 (0)	0.2323	0.4971	physionet.org, Y. Chen & E. Keogh
23	Device	ElectricDevices	8926	7711	7	96	0.4492	0.3806 (14)	0.3988	0.7463	A. Bagnall & J. Lines
24	Image	FaceAll	560	1690	14	131	0.2864	0.1917 (3)	0.1923	0.8302	X. Xi & E. Keogh
25	Image	FaceFour	24	88	4	350	0.2159	0.1136 (2)	0.1705	0.7045	A. Ratanamahatana & E. Keogh
26	Image	FacesUCR	200	2050	14	131	0.2307	0.0878 (12)	0.0951	0.8566	X. Xi & E. Keogh
27	Image	FiftyWords	450	455	50	270	0.3692	0.2418 (6)	0.3099	0.8747	T. Rath & R. Manmatha
28	Image	Fish	175	175	7	463	0.2171	0.1543 (4)	0.1771	0.8343	D. Lee
29	Sensor	FordA	3601	1320	2	500	0.3348	0.3091 (1)	0.4455	0.4841	A. Bagnall
30	Sensor	FordB	3636	810	2	500	0.3938	0.3926 (1)	0.3802	0.4951	A. Bagnall
31	Motion	GunPoint	50	150	2	150	0.0867	0.0867 (0)	0.0933	0.4933	A. Ratanamahatana & E. Keogh
32	Spectro	Ham	109	105	2	431	0.4000	0.4000 (0)	0.5333	0.4857	K. Kemsley & A. Bagnall
33	Image	HandOutlines	1000	370	2	2709	0.1378	0.1378 (0)	0.1189	0.3595	L. Davis & A. Bagnall
34	Motion	Haptics	155	308	5	1092	0.6299	0.5877 (2)	0.6234	0.7825	J. Brady
35	Image	Herring	64	64	2	512	0.4844	0.4688 (5)	0.4688	0.4063	J. Maap & A. Bagnall
36	Motion	InlineSkate	100	550	7	1882	0.6582	0.6127 (14)	0.6164	0.8164	F. Morchen & O. Hoos
37	Sensor	InsectWingbeatSound	220	1980	11	256	0.4384	0.4152 (1)	0.6449	0.9091	Y. Chen & E. Keogh
38	Sensor	ItalyPowerDemand	67	1029	2	24	0.0447	0.0447 (0)	0.0496	0.4985	JJ van Wijk, E. Keogh & L. Wi
39	Device	LargeKitchenAppliances	375	375	3	720	0.5067	0.2053 (94)	0.2053	0.6667	J. Lines & A. Bagnall
40	Sensor	Lightning2	60	61	2	637	0.2459	0.1311 (6)	0.1311	0.4590	D. Eads
41	Sensor	Lightning7	70	73	7	319	0.4247	0.2877 (5)	0.2740	0.7397	D. Eads
42	Simulated	Mallat	55	2345	8	1024	0.0857	0.0857 (0)	0.0661	0.8729	M. Jeong & S. Mallat
43	Spectro	Meat	60	60	3	448	0.0667	0.0667 (0)	0.0667	0.6667	K. Kemlsey & A. Bagnall
44	Image	MedicalImages	381	760	10	99	0.3158	0.2526 (20)	0.2632	0.4855	J. Felipe & C. Traina
45	Image	MiddlePhalanxOutlineAgeGroup	400	154	3	80	0.4805	0.4805 (0)	0.5000	0.4286	L. Davis & A. Bagnall
46	Image	MiddlePhalanxOutlineCorrect	600	291	2	80	0.2337	0.2337 (0)	0.3024	0.4296	L. Davis & A. Bagnall
47	Image	MiddlePhalanxTW	399	154	6	80	0.4870	0.4935 (3)	0.4935	0.7143	L. Davis & A. Bagnall
48	Sensor	MoteStrain	20	1252	2	84	0.1214	0.1342 (1)	0.1653	0.4609	C. Guestrin & J. Sun
49	ECG	NonInvasiveFetalECGThorax1	1800	1965	42	750	0.1710	0.1893 (1)	0.2097	0.9705	physionet.org, B. Hu & E. Keogh
50	ECG	NonInvasiveFetalECGThorax2	1800	1965	42	750	0.1201	0.1290 (1)	0.1354	0.9705	physionet.org, B. Hu & E. Keogh
51	Spectro	OliveOil	30	30	4	570	0.1333	0.1333 (0)	0.1667	0.6000	K. Kemsley & A. Bagnall
52	Image	OSULeaf	200	242	6	427	0.4793	0.3884 (7)	0.4091	0.7727	A. Gandhi
53	Image	PhalangesOutlinesCorrect	1800	858	2	80	0.2389	0.2389 (0)	0.2716	0.3869	A. Bagnall
54	Sensor	Phoneme	214	1896	39	1024	0.8908	0.7727 (14)	0.7716	0.8871	H. Hamooni & A. Mueen
55	Sensor	Plane	105	105	7	144	0.0381	0.0000 (5)	0.0000	0.8000	J. Gao
56	Image	ProximalPhalanxOutlineAgeGroup	400	205	3	80	0.2146	0.2146 (0)	0.1951	0.5122	L. Davis & A. Bagnall
57	Image	ProximalPhalanxOutlineCorrect	600	291	2	80	0.1924	0.2096 (1)	0.2165	0.3162	L. Davis & A. Bagnall
58	Image	ProximalPhalanxTW	400	205	6	80	0.2927	0.2439 (2)	0.2439	0.6488	L. Davis & A. Bagnall
59	Device	RefrigerationDevices	375	375	3	720	0.6053	0.5600 (8)	0.5360	0.6667	J. Lines & A. Bagnall
60	Device	ScreenType	375	375	3	720	0.6400	0.5893 (17)	0.6027	0.6667	J. Lines & A. Bagnall
61	Simulated	ShapeletSim	20	180	2	500	0.4611	0.3000 (3)	0.3500	0.5000	J. Hills & A. Bagnall
62	Image	ShapesAll	600	600	60	512	0.2483	0.1980 (4)	0.2317	0.9833	J. Hills & A. Bagnall
63	Device	SmallKitchenAppliances	375	375	3	720	0.6587	0.3280 (15)	0.3573	0.6667	J. Lines & A. Bagnall
64	Sensor	SonyAIBORobotSurface1	20	601	2	70	0.3045	0.3045 (0)	0.2745	0.4293	D. Vail, M. Velso & E. Keogh
65	Sensor	SonyAIBORobotSurface2	27	953	2	65	0.1406	0.1406 (0)	0.1689	0.3830	D. Vail, M. Velso & E. Keogh
66	Sensor	StarLightCurves	1000	8236	3	1024	0.1512	0.0947 (16)	0.0934	0.4228	P. Protopapas, E. Keogh & L. Wei
67	Spectro	Strawberry	613	370	2	235	0.0541	0.0541 (0)	0.0595	0.3568	K. Kemsley & A. Bagnall
68	Image	SwedishLeaf	500	625	15	128	0.2112	0.1536 (2)	0.2080	0.9216	O. Soderkvist
69	Image	Symbols	25	995	6	398	0.1005	0.0623 (8)	0.0503	0.8211	E. Keogh & J. Brady
70	Simulated	SyntheticControl	300	300	6	60	0.1200	0.0167 (6)	0.0067	0.8333	R. Alcock & Y. Manolopoulos
71	Motion	ToeSegmentation1	40	228	2	277	0.3202	0.2500 (8)	0.2281	0.4737	A. Bagnall, L. Ye & E. Keogh
72	Motion	ToeSegmentation2	36	130	2	343	0.1923	0.0923 (5)	0.1615	0.1846	A. Bagnall, L. Ye & E. Keogh
73	Sensor	Trace	100	100	4	275	0.2400	0.0100 (3)	0.0000	0.7100	D. Roverso
74	ECG	TwoLeadECG	23	1139	2	82	0.2529	0.1317 (4)	0.0957	0.4996	physionet.org & E. Keogh
75	Simulated	TwoPatterns	1000	4000	4	128	0.0932	0.0015 (4)	0.0000	0.7412	P. Geurts
76	Motion	UWaveGestureLibraryAll	896	3582	8	945	0.0519	0.0343 (4)	0.1083	0.8716	A. Bagnall & J. Liu
77	Motion	UWaveGestureLibraryX	896	3582	8	315	0.2607	0.2267 (4)	0.2725	0.8716	J. Liu
78	Motion	UWaveGestureLibraryY	896	3582	8	315	0.3384	0.3009 (4)	0.3660	0.8716	J. Liu
79	Motion	UWaveGestureLibraryZ	896	3582	8	315	0.3504	0.3222 (6)	0.3417	0.8716	J. Liu
80	Sensor	Wafer	1000	6164	2	152	0.0045	0.0045 (1)	0.0201	0.1079	R. Olszewski
81	Spectro	Wine	57	54	2	234	0.3889	0.3889 (0)	0.4259	0.5000	K. Kemsley & A. Bagnall
82	Image	WordSynonyms	267	638	25	270	0.3824	0.2618 (9)	0.3511	0.7806	T. Rath & R. Manmatha
83	Motion	Worms	181	77	5	900	0.5455	0.4675 (9)	0.4156	0.5714	A. Bagnall
84	Motion	WormsTwoClass	181	77	2	900	0.3896	0.4156 (7)	0.3766	0.4286	A. Bagnall
85	Image	Yoga	300	3000	2	426	0.1697	0.1560 (7)	0.1637	0.4643	L. Wei & E. Keogh
86	Device	ACSF1	100	100	10	1460	0.4600	0.3800 (4)	0.3600	0.9000	P. Schafer
87	Sensor	AllGestureWiimoteX	300	700	10	Vary	0.4843	0.2829 (14)	0.2843	0.9000	J. Guna
88	Sensor	AllGestureWiimoteY	300	700	10	Vary	0.4314	0.2700 (9)	0.2714	0.9000	J. Guna
89	Sensor	AllGestureWiimoteZ	300	700	10	Vary	0.5457	0.3486 (11)	0.3571	0.9000	J. Guna
90	Simulated	BME	30	150	3	128	0.1667	0.0200 (4)	0.1000	0.6667	Joseph Fourier University
91	Traffic	Chinatown	20	343	2	24	0.0466	0.0466 (0)	0.0437	0.2741	H.A. Dau
92	Image	Crop	7200	16800	24	46	0.2883	0.2883 (0)	0.3348	0.9583	F. Petitjean
93	Sensor	DodgerLoopDay	78	80	7	288	0.4500	0.4125 (1)	0.5000	0.8375	C.-C. M. Yeh
94	Sensor	DodgerLoopGame	20	138	2	288	0.1159	0.0725 (1)	0.1232	0.4783	C.-C. M. Yeh
95	Sensor	DodgerLoopWeekend	20	138	2	288	0.0145	0.0217 (1)	0.0507	0.2609	C.-C. M. Yeh
96	EOG	EOGHorizontalSignal	362	362	12	1250	0.5829	0.5249 (1)	0.4972	0.9144	E. Keogh & H. A. Dau
97	EOG	EOGVerticalSignal	362	362	12	1250	0.5580	0.5249 (2)	0.5525	0.9144	E. Keogh & H. A. Dau
98	Spectro	EthanolLevel	504	500	4	1751	0.7260	0.7180 (1)	0.7240	0.7480	A. Bagnall
99	Sensor	FreezerRegularTrain	150	2850	2	301	0.1951	0.0930 (1)	0.1011	0.5000	REFIT project
100	Sensor	FreezerSmallTrain	28	2850	2	301	0.3242	0.3242 (0)	0.2411	0.5000	REFIT project
101	HRM	Fungi	18	186	18	201	0.1774	0.1774 (0)	0.1613	0.8978	W. Fonzi
102	Trajectory	GestureMidAirD1	208	130	26	Vary	0.4231	0.3615 (5)	0.4308	0.9615	H. A. Dau
103	Trajectory	GestureMidAirD2	208	130	26	Vary	0.5077	0.4000 (6)	0.3923	0.9615	H. A. Dau
104	Trajectory	GestureMidAirD3	208	130	26	Vary	0.6538	0.6231 (1)	0.6769	0.9615	H. A. Dau
105	Sensor	GesturePebbleZ1	132	172	6	Vary	0.2674	0.1744 (2)	0.2093	0.8140	I. Maglogiannis
106	Sensor	GesturePebbleZ2	146	158	6	Vary	0.3291	0.2215 (6)	0.3291	0.8101	I. Maglogiannis
107	Motion	GunPointAgeSpan	135	316	2	150	0.1013	0.0348 (3)	0.0823	0.4937	A. Ratanamahatana & E. Keogh
108	Motion	GunPointMaleVersusFemale	135	316	2	150	0.0253	0.0253 (0)	0.0032	0.4747	A. Ratanamahatana & E. Keogh
109	Motion	GunPointOldVersusYoung	136	315	2	150	0.0476	0.0349 (4)	0.1619	0.4762	A. Ratanamahatana & E. Keogh
110	Device	HouseTwenty	40	119	2	2000	0.3361	0.0588 (33)	0.0756	0.4202	E. Keogh & S. Gharghabi
111	EPG	InsectEPGRegularTrain	62	249	3	601	0.3213	0.1727 (11)	0.1285	0.5261	E. Keogh & S. Gharghabi
112	EPG	InsectEPGSmallTrain	17	249	3	601	0.3373	0.3052 (1)	0.2651	0.5261	E. Keogh & S. Gharghabi
113	Traffic	MelbournePedestrian	1194	2439	10	24	0.1525	0.1845 (1)	0.2091	0.8995	H.A. Dau
114	Image	MixedShapesRegularTrain	500	2425	5	1024	0.1027	0.0911 (4)	0.1584	0.7303	E. Keogh
115	Image	MixedShapesSmallTrain	100	2425	5	1024	0.1645	0.1674 (7)	0.2202	0.7303	E. Keogh
116	Sensor	PickupGestureWiimoteZ	50	50	10	Vary	0.4400	0.3400 (17)	0.3400	0.9000	J. Guna
117	Hemodynamics	PigAirwayPressure	104	208	52	2000	0.9423	0.9038 (1)	0.8942	0.9808	M. Guillame-Bert
118	Hemodynamics	PigArtPressure	104	208	52	2000	0.8750	0.8029 (1)	0.7548	0.9808	M. Guillame-Bert
119	Hemodynamics	PigCVP	104	208	52	2000	0.9183	0.8413 (11)	0.8462	0.9808	M. Guillame-Bert
120	Device	PLAID	537	537	11	Vary	0.4767	0.1657 (12)	0.1639	0.8380	P. Schafer
121	Power	PowerCons	180	180	2	144	0.0667	0.0778 (3)	0.1222	0.5000	EDF R&D, France
122	Spectrum	Rock	20	50	4	2844	0.1600	0.1600 (0)	0.4000	0.5800	Y. Zhu
123	Spectrum	SemgHandGenderCh2	300	600	2	1500	0.2383	0.1550 (1)	0.1983	0.3500	C.-C. M. Yeh
124	Spectrum	SemgHandMovementCh2	450	450	6	1500	0.6311	0.3622 (1)	0.4156	0.8333	C.-C. M. Yeh
125	Spectrum	SemgHandSubjectCh2	450	450	5	1500	0.5956	0.2000 (3)	0.2733	0.8000	C.-C. M. Yeh
126	Sensor	ShakeGestureWiimoteZ	50	50	10	Vary	0.4000	0.1600 (6)	0.1400	0.9000	J. Guna
127	Simulated	SmoothSubspace	150	150	3	15	0.0933	0.0533 (1)	0.1733	0.6667	X. Huang
128	Simulated	UMD	36	144	3	150	0.2361	0.0278 (6)	0.0069	0.6667	Joseph Fourier University

数据集下载后文件夹中已经分割了训练和测试集，就是上图中的Train和Test，Class代表了这个数据集包含了几类，Length代表了数据集样本的长度（每个数据集中所有序列都是等长的）。后面的ED代表了欧式距离，DTW就是动态时间规整算法，DTW（learned_w）代表了窗口约束DTW算法（DTW的改进算法）。这几个算法都是空间距离度量算法，下面的误差率是在1-NN分类算法下，结合ED、DTW及其改进距离度量算法的分类误差率。相关文章很多，可以见下面这篇期刊：

---

---

---

参考资料：
UCR Time Series Classification Archive 是个什么数据集?
超全必看！开源时间序列数据集整理