17.2.4 Nonlinear Principal Components Analysis
Probably the most well-known method to perform dimension reduction and, thus,
for creating maps is principal components analysis (PCA). PCA has a numerical
data matrix as input and linearly transforms the data, such that the first dimension,
normally plotted in horizontal direction, explains the variance in the data as much as
possible. The next dimensions (in case of a map only the second) try to do the same
for the remaining variance. Also, all dimensions are uncorrelated with each other.
Nonlinear principal components analysis (NL-PCA) [13, 31, 34] is a method that
does the same as PCA, but is also able to handle categorical attributes and missing
values. Also, using ordinal transformation, numerical attributes can be transformed
nonlinearly. Additionally, the categories of the attributes also have a location in the
map, that is in the center of the items belonging to that category. These category
17.2.4 Nonlinear Principal Components AnalysisProbably the most well-known method to perform dimension reduction and, thus,for creating maps is principal components analysis (PCA). PCA has a numericaldata matrix as input and linearly transforms the data, such that the first dimension,normally plotted in horizontal direction, explains the variance in the data as much aspossible. The next dimensions (in case of a map only the second) try to do the samefor the remaining variance. Also, all dimensions are uncorrelated with each other.Nonlinear principal components analysis (NL-PCA) [13, 31, 34] is a method thatdoes the same as PCA, but is also able to handle categorical attributes and missingvalues. Also, using ordinal transformation, numerical attributes can be transformednonlinearly. Additionally, the categories of the attributes also have a location in themap, that is in the center of the items belonging to that category. These category
การแปล กรุณารอสักครู่..
