|SWAG >> Conferences >> UCC 2014|
Ian J. Davis,
Michael W. Godfrey,
Douglas Neuse, Serge Mankovskii
This code implements the greedy algorithm proposed to detect anomalies in timeseries. It does this by finding the data points in the time series which are least smooth.
This source code has been compiled on a windows platform, but should be easy to port to any platform. It involves only one short file.
This code takes directories and/or file names as input. It scans directories recursively for file inputs, permitting large tests to be conducted on a host of files easily.
Input format are double precision values, separated by commas and/or new lines. Line comments are identified by a # at the start of a line, and otherwise ignored.
Output (in a format suitable for Excel) consists of the input values, the computed cost using the techniques described in the above paper, and the normalized cost.
For more information on this research please contact us at .