COMPARATIVE ANALYSIS OF MISSING DATA RECOVERY METHODS

Abstract

In recent decades, the methods of system analysis have been developing qualitatively. It is
associated with an increase in the rate of technical development, the densification of time processes,
the rapid growth of accumulated information and new capabilities of computer technology.
These include methods for analyzing large amounts of data, methods of data mining, methods of
analytical modeling, methods of parallel data processing, neural network methods, forecasting
methods, and others. The presented methods make it possible to quickly and efficiently process
heterogeneous clusters of information, accumulate and synthesize data, generalize and classify
information. The last of the presented methods are methods of interpolation and extrapolation of
lost, damaged or missing information. These methods allow to structure, restore and model information
based on statistical data, mathematical and algorithmic methods. Thus, the article deals
with the problem of recovering missing data in graphic and complex objects. Literary sources on
the problems under consideration are given. They provide extensive information on the topic under
consideration: present genetic algorithms used for spatial interpolation; the solution of problems
of heterogeneity of interpolation of seismic data is considered; it is described the use of
spline approximation to calculate the characteristics of nonlinear electronic components; the
method of constructing a model of three-dimensional parametric rational bodies using generalized
Bezier interpolation is analyzed, which allows modeling the shape of a body and anisotropic
space; methods using fuzzy linear equations are described, which are widespread in computer
vision; the method of adaptive interpolation based on the gradient and taking into account the
local gradient of the original image is investigated. It is made comparing several common methods
of interpolation and data restoration, in article, such as: bilinear interpolation, Bezier surface.
Each method and features of its application within the framework of the experiment are briefly
described. The result of a series of experiments with the presented methods with different numbers
of tests is presented. In conclusion, summary is drawn about the rationality of choosing one of the
proposed methods without the use of a long field experiment in each case.