Is there any way to use eigenvalues to filter and classify the sample?

I can obtain the factor with the highest eigenvalue by using SVM or random forest. Is there any way to use eigenvalues to filter and classify the sample? For example, I have 3 factors X, Y, and Z. All samples have these three attributes, and all samples have classification C. I want to classify samples according to the contribution degree of X, Y, Z to classification C.