Principal component analysis (PCA) PCA is a dimension reduction techniques. So, if we have a large number of predictors, instead of using all the predictors for modelling or other analysis, we can compressed all the information from the variables and create a new set of variables.