22
Volume 5, Issue 10: Special Issue
(EJAR)
ISSN: 2181-2020
MPHAPP
THE 6TH INTERNATIONAL SCIENTIFIC AND PRACTICAL
CONFERENCE
“
MODERN PHARMACEUTICS: ACTUAL
PROBLEMS AND PROSPECTS
”
TASHKENT, OCTOBER 17, 2025
in-academy.uz
METHODS FOR ANALYZING MEDICAL AND PHARMACEUTICAL DATA
PROVIDED IN NON-QUANTITATIVE FORM
Tursunov A.T.
1
Narzullayev D.Z.
2
Rahmonov E.D.
3
Tashkent Pharmaceutical Institute, Tashkent city, Republic of Uzbekistan
e-mail: tursunovr484z @ gmail.com.ru
https://doi.org/10.5281/zenodo.17311693
Relevance:
The heuristic approach is based on the researcher's knowledge and intuition, which
are difficult to formalize. In this approach, the researcher determines what information and how to
use it to achieve the required level of recognition efficiency. Therefore, when analyzing medical and
pharmaceutical data presented in quantitative form, it is relevant to convert them into a single type,
that is, into a digital form.
Research Objective.
The two fundamental methods of knowledge representation described
above allow us to propose the following classification of pattern recognition methods. This requires
new approaches to analyzing large volumes of medical and pharmaceutical data. To achieve the goal,
we consider the following methods.
Method and Methods.
Intentional methods of symbol recognition - methods based on
performing operations with symbols. Extensional methods of symbol recognition - methods based on
performing operations with objects.
At the same time, objects in these methods are not considered as whole information units, but
rather as indicators for evaluating the interaction of their attributes and behavior. The group of
intensional methods of pattern recognition is very wide, and its division into subclasses is, in a sense,
conditional. These methods are used to determine the ratio of similarities in different areas of a
multidimensional feature space. This group also includes a method for calculating the likelihood ratio
for unrelated traits. This method does not require knowledge of the functional form of the distribution
law, except for the assumption that the traits are unrelated (which is almost never true in reality).
Therefore, it can be classified as nonparametric. Other nonparametric methods are of particular
interest when the shape of the distribution density curve is unknown and no assumptions are made
about its nature. These include the method of multidimensional histograms, the method of “k–nearest
neighbors”, the method of Euclidean distance, the method of potential functions, etc., a generalization
of which is the method of “Parzen estimates”. These methods formally work with objects as whole
structures, but depending on the type of recognition problem, they can be either intensional or
extensional forms.
Results.
Nonparametric methods analyze the relative number of objects in a given
multidimensional volume and use different distance functions between the objects in the training
sample and the objects being recognized. For quantitative features, when their number is much
smaller than the sample size, operations with objects play an intermediate role in estimating the local
density of conditional probabilities.
Conclusions.
Thus, the same technological operations of nonparametric methods, depending
on the conditions of the problem, have the meaning of locally estimating the density of the probability
distribution of character values or assessing the similarity and difference of objects. When classifying
an unknown object, a given number (k) of other objects (nearest neighbors) known to belong to the
classes being recognized are found in the feature space.
