Using Kernel Principal Component Analysis of Treating Linear Multiplicity Problem for Covid-19 injured Data

Abstract

The main goal of using multivariate analysis method is to summarize the large number of data whose explanatory variables are correlated with each other by strong and complex relationships, i.e. most of them are simplistic methods that help the researcher form an idea and conclusion about these overlapping groups. The classical principal component analysis method is used to turn a set of related variables into orthogonal components called principal components, but it is difficult to deal with this data in the principal component method if the data matrix is nonlinear. We used (18) variables representing the governorates of Iraq and data about the number of people infected with the new Coronavirus, based on the daily epidemiological situation of the Public Health Department of the Iraqi Ministry of Health This research aims to use Kernel Principal Component Analysis (KPCA) to deal with the Non-linear data set. It is similar to Principal Component Analysis but it's mapping the data in high dimensional space called feature space. The results show that the problem of linear multiplicity can be addressed by using principal component analysis method, where the correlated variables represented with a smaller number of orthogonal components, which amounted to (14) principal component that explained a percentage (84%) of the total variance.