This study addresses the challenge of detecting coarse
measurements (outliers) in time series data, a common issue in fields such as
space geodynamics, geodesy, and other measurement-driven sciences. The proposed
outlier detection algorithm solves two key problems: first, it finds a solution
containing the maximum possible amount of measurement data remaining after
detecting and removing outliers. Second, it requires the minimum possible
number of arithmetic operations, estimated by the value O(NlogN), which cannot
be improved in order N, to find a solution. To construct the algorithm, it was
shown that the required solution should be sought in the ordered sequence of
input data in the form of a set of consecutive numbers. The search for the set
of maximum length is performed step by step, with the range of possible values
of the set length being halved at each step.
The transition to one of the two halves is performed by checking the
fulfilment of certain criteria, not for all possible values of lengths from the
range under consideration, but only for its mid value. Testing the algorithm on
real data obtained from laser rangefinder measurements showed the advantages of
it over others, both in terms of time consumption and the amount of data
remaining after detection and removal of outliers. The algorithm is robust and
always finds a solution, if one exists. Its time efficiency is evident when
cleaning a large amount of measurement data of outliers. It can be used for
automated cleaning from outliers of observation data in information and
measuring systems, in systems with artificial intelligence, as well as when
solving various scientific, applied managerial and other problems using modern
computer systems in order to obtain promptly the most accurate final result.
Author(s) Details
Igor V.Bezmenov
Russian Metrological Institute of Technical Physics and Radio Engineering,
Mendeleevo, Moscow Region, Russia.
Please see the link:- https://doi.org/10.9734/bpi/psniad/v3/6424
No comments:
Post a Comment