Database Reference
In-Depth Information
30
25
20
0123456789 0 1 2 3 4 5
time
Fig. 1.
Example of a time series.
Table 1.
Data sets used in the experiments.
Data set
Description
Number of
Points
Total number
series
per series
of points
Stock prices
98 stocks, 2.3 years
98
610
60,000
Air and sea
68 buoys, 18 years,
136
1,800-6,600
450,000
Temperatures
2 sensors per buoy
Wind speeds
12 stations, 18 years
12
6,570
79,000
EEG
64 electrodes, 1 second
64
256
16,000
series and show that it works well with compressed data. Finally, we present
a technique for indexing and retrieval of compressed series; we have tested
it on four data sets (Table 1), which are publicly available through the
Internet.
Stock prices:
We have used stocks from the Standard and Poor's 100
listing of large companies for the period from January 1998 to April 2000.
We have downloaded daily prices from America Online, discarded newly
listed and de-listed stocks, and used ninety-eight stocks in the experiments.
Air and sea temperatures:
We have experimented with daily temperature
readings by sixty-eight buoys in the Pacific Ocean, from 1980 to 1998,
downloaded from the Knowledge Discovery archive at the University of
California at Irvine (kdd.ics.uci.edu).
Wind speeds:
We have used daily wind speeds from twelve sites in
Ireland, from 1961 to 1978, obtained from an archive at Carnegie Mellon
University (lib.stat.cmu.edu/datasets).
Electroencephalograms:
We have utilized EEG obtained by Henri
Begleiter at the Neurodynamics Laboratory of the SUNY Health Center
at Brooklyn. These data are from sixty-four electrodes at standard points
on the scalp; we have downloaded them from an archive at the University
of California at Irvine (kdd.ics.uci.edu).
