We figured out the implementation of the PCA algorithm using the numpy library. Scikit-learn can let us start using this method with just one line of code:

from sklearn.decomposition import PCA
pca_model = PCA(n_components = 2)

PCA is a scikit-learn library class. It contains more than 5 arguments, but we are most interested in only one - n_components. This argument is responsible for the number of main components that we want to get. The only condition is that the number of components must, of course, be equal to or less than the variables in the dataset. The PCA class contains 2 main methods that we will use: fit and transform. The fit() method loads the data into the class, and the transform() method transforms it, and we get the result of the PCA algorithm. If we want to combine these 2 operations, use the fit_transform() method:

pca_model = PCA(n_components = 2)

# fit() and transform()
pca_model.fit(X)
X_reduced = pca_model.transform(X)

# fit_transform()
X_reduced = pca_model.fit_transform(X)

If we want to get the components that the algorithm has calculated, call the .components_ attribute:

print(pca_model.components_)