Powered by OpenAIRE graph
Found an issue? Give us feedback
addClaim

This Research product is the result of merged Research products in OpenAIRE.

You have already added 0 works in your ORCID record related to the merged Research product.

Поиск выбросов в Ð¼Ð½Ð¾Ð³Ð¾Ð¼ÐµÑ€Ð½Ñ‹Ñ Ð´Ð°Ð½Ð½Ñ‹Ñ , основанный на SV-Ð¿Ð¾Ð´Ñ Ð¾Ð´Ðµ

выпускная квалификационная работа бакалавра

Поиск выбросов в Ð¼Ð½Ð¾Ð³Ð¾Ð¼ÐµÑ€Ð½Ñ‹Ñ Ð´Ð°Ð½Ð½Ñ‹Ñ , основанный на SV-Ð¿Ð¾Ð´Ñ Ð¾Ð´Ðµ

Abstract

В работе рассматриваются методы поиска выбросов в многомерных данных. Основное внимание уделяется алгоритмам, основанным на SV-подходе – SVC и OneClass SVM. Приводится сравнение результатов работы этих алгоритмов с результатами метода, основанного на расстоянии Махаланобиса. Задачи, которые решались в ходе исследования: 1. Изучение отобранных алгоритмов поиска выбросов и их реализация. 2. Тестирование реализации на модельных данных и их апробация на реальных данных двух типов. 3. Анализ результатов. В процессе работы были проведены численные эксперименты, показывающие наглядно, каким образом параметры алгоритмов влияют на эффективность выявления выбросов. Оригинальные коды, реализующие алгоритмы, были разработаны в среде Google Colab с использованием языка программирования Python. Реальные наборы данных были взяты из открытых источников. Был проведен сравнительный анализ работы трех алгоритмов обнаружения выбросов, основанный на результатах, полученных при проведении численных экспериментов. Разработаны методологические рекомендации по использованию алгоритмов для выявления выбросов в многомерных данных. На основании проведенных исследований на реальных данных были сделаны выводы о возможности эффективного применения 5 алгоритмов при анализе многомерных данных различного типа. Исследования, проведенные на реальных данных с применением SV-алгоритмов, продемонстрировали возможность построения эффективной SV-машины для обнаружения выбросов по тренировочным данным с последующим применением ее для обнаружения выбросов в новых данных той же природы.

This work examines methods for detecting outliers in multidimensional data. The main focus is on algorithms based on the SV approach – SVC and OneClass SVM. A comparison of the performance of these algorithms with the results of the method based on the Mahalanobis distance is presented. The tasks addressed during the research are: 1. Study of the selected outlier detection algorithms and their implementation. 2. Testing the implementation on model data and its validation on two types of real data. 3. Analysis of the results. Experiments were conducted in the paper to visually demonstrate how the parameters of the algorithms affect the efficiency of outlier detection. The analysis of the methods was carried out using software equipped with the necessary statistical functions and methods. The original codes implementing the algorithms were developed in the Google Colab environment using the Python programming language. Real datasets were sourced from open access. A comparative analysis of the performance of the three outlier detection algorithms was carried out based on the results obtained from numerical experiments. Methodological recommendations for using the algorithms to detect outliers in multidimensional data were developed. Based on the research conducted on real data, conclusions were drawn about the possibility of effectively applying the algorithms when analyzing multidimensional data of various natures.

Keywords

test set, расстояние ÐœÐ°Ñ Ð°Ð»Ð°Ð½Ð¾Ð±Ð¸ÑÐ°, SVC, outliers, опорные векторы, выбросы в Ð¼Ð½Ð¾Ð³Ð¾Ð¼ÐµÑ€Ð½Ñ‹Ñ Ð´Ð°Ð½Ð½Ñ‹Ñ, тестовая выборка, Mahalanobis distance, training set, OneClass SVM, тренировочная выборка

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Beta
sdg_colorsSDGs:
Related to Research communities
Upload OA version
Are you the author of this publication? Upload your Open Access version to Zenodo!
It’s fast and easy, just two clicks!