Robust Principal Component Analysis Based on Modified Minimum Covariance Determinant in the Presence of Outliers

B. Barış Alkan

Alphanumeric Journal, 2016, vol. 4, issue 2, 85-94

Abstract: Principal component analysis (PCA) is not resistant to outliers existing in multivariate data sets. The results which are obtained by using classical PCA are far from real values in the presence of outliers. Therefore, using robust versions of PCA is favorable. The easiest way to obtain robust principal components is to replace classical estimates of the location and scale parameters with their robust versions. Robust estimations of location and scale parameters can be found with minimum covariance determinant (MCD) providing high breakdown point. In this study, algorithm of MCD is modified using Jackknife resampling approach and results of this modification are examined. Proposed robust principal component analysis (RPCA) based on modified MCD (MMCD) method that is modified using Jaccknife resampling are evaluated over two real data with different outlier ratios. In the light of obtained results, it can be said that RPCA based on MMCD is better than RPCA based on MCD in the presence of outliers.

Keywords: Minimum Covariance Determinant; Outliers; Robust Principal Component Analysis (search for similar items in EconPapers)
JEL-codes: C38 (search for similar items in EconPapers)
Date: 2016
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.alphanumericjournal.com/media/Issue/vo ... varyans-_5YcrjpT.pdf (application/pdf)
https://alphanumericjournal.com/article/aykiri-goz ... -bilesenler-analizi/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:anm:alpnmr:v:4:y:2016:i:2:p:85-94

DOI: 10.17093/aj.2016.4.2.5000189525

Access Statistics for this article

More articles in Alphanumeric Journal from Bahadir Fatih Yildirim
Bibliographic data for series maintained by Bahadir Fatih Yildirim ().