Principle Component Analysis Based on New Symmetric Similarity Measures for Heavy-Tailed Data
Keyword(s):
We consider the principal component analysis (PCA) for the heavy-tailed distributions. A traditional measure for the classical PCA is the covariance measure. Due to the non-existence of variance of many heavy-tailed distributions, this measure cannot be used for them. We will clarify how to perform PCA in heavy-tailed data by extending a similarity measure based on covariance. We introduce similarity measures based on a new dependence coefficient of heavy-tailed distributions. Using real and artificial datasets, the performance of the proposed PCA is evaluated and compared with the classical one.
2021 ◽
Vol 23
(06)
◽
pp. 1699-1715
2014 ◽
Vol 635-637
◽
pp. 997-1000
◽
2000 ◽
Vol 92
(6)
◽
pp. 1545-1552
◽
2017 ◽
Vol 2017
◽
pp. 1-8
◽
2021 ◽
Vol 12
(2)
◽
pp. 835-844
2019 ◽
Vol 20
(3)
◽
pp. 119
2018 ◽
Vol 7
(2.32)
◽
pp. 233