A good PCA will be at approximately 80%. If PCA is a weaker component, you are losing data.
Using the cnbc dataset and brand_data, I am:
a)Preparing a Data Set for PCA
b)Conducting PCA
c)Selecting Principal Components
d)Building a Predictive Model Using Principal Components
e)Making Out of Sample Predictions Using Principal Components
f)Examine PCA Performance relative to an OLS model
g)Building a Classification Model Using Principal Components
h)Making Out of Sample Predictions Using Principal Components
i)Examine PCA Performance relative to a classification (Naive Bayes) model
I'm also using the Boston Housing Values dataset (Boston) in the package MASS. It contains data on the value of owner-occupied homes and its determinants.