Dimensionality Reduction using PCA

Dimensionality reduction, an unsupervised machine learning method is used to reduce the number of feature variables for each data sample selecting set of principal features. Principal Component Analysis (PCA) is one of the popular algorithms for dimensionality reduction.

Exact PCA

Principal Component Analysis (PCA) is used for linear dimensionality reduction using Singular Value Decomposition (SVD) of the data to project it to a lower dimensional space. While decomposition using PCA, input data is centered but not scaled for each feature before applying the SVD.

The Scikit-learn ML library provides sklearn.decomposition.PCA module that is implemented as a transformer object which learns n components in its fit() method. It can also be used on new data to project it on these components.

Example

The below example will use sklearn.decomposition.PCA module to find best 5 Principal components from Pima Indians Diabetes dataset.

from pandas import read_csv
from sklearn.decomposition import PCA
path = r'C:\Users\Leekha\Desktop\pima-indians-diabetes.csv'
names = ['preg', 'plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', ‘class']
dataframe = read_csv(path, names = names)
array = dataframe.values
X = array[:,0:8]
Y = array[:,8]
pca = PCA(n_components = 5)
fit = pca.fit(X)
print(("Explained Variance: %s") % (fit.explained_variance_ratio_))
print(fit.components_)

Output

Explained Variance: [0.88854663 0.06159078 0.02579012 0.01308614 0.00744094]
[
   [-2.02176587e-03 9.78115765e-02 1.60930503e-02 6.07566861e-029.93110844e-01 1.40108085e-02 5.37167919e-04 -3.56474430e-03]
   [-2.26488861e-02 -9.72210040e-01 -1.41909330e-01 5.78614699e-029.46266913e-02 -4.69729766e-02 -8.16804621e-04 -1.40168181e-01]
   [-2.24649003e-02 1.43428710e-01 -9.22467192e-01 -3.07013055e-012.09773019e-02 -1.32444542e-01 -6.39983017e-04 -1.25454310e-01]
   [-4.90459604e-02 1.19830016e-01 -2.62742788e-01 8.84369380e-01-6.55503615e-02 1.92801728e-01 2.69908637e-03 -3.01024330e-01]
   [ 1.51612874e-01 -8.79407680e-02 -2.32165009e-01 2.59973487e-01-1.72312241e-04 2.14744823e-02 1.64080684e-03 9.20504903e-01]
]

Incremental PCA

Incremental Principal Component Analysis (IPCA) is used to address the biggest limitation of Principal Component Analysis (PCA) and that is PCA only supports batch processing, means all the input data to be processed should fit in the memory.

The Scikit-learn ML library provides sklearn.decomposition.IPCA module that makes it possible to implement Out-of-Core PCA either by using its partial_fit method on sequentially fetched chunks of data or by enabling use of np.memmap, a memory mapped file, without loading the entire file into memory.

Same as PCA, while decomposition using IPCA, input data is centered but not scaled for each feature before applying the SVD.

Example

The below example will use sklearn.decomposition.IPCA module on Sklearn digit dataset.

from sklearn.datasets import load_digits
from sklearn.decomposition import IncrementalPCA
X, _ = load_digits(return_X_y = True)
transformer = IncrementalPCA(n_components = 10, batch_size = 100)
transformer.partial_fit(X[:100, :])
X_transformed = transformer.fit_transform(X)
X_transformed.shape

Output

(1797, 10)

Here, we can partially fit on smaller batches of data (as we did on 100 per batch) or you can let the fit() function to divide the data into batches.

Explore our latest online courses and learn new skills at your own pace. Enroll and become a certified expert to boost your career.

Kernel PCA

Kernel Principal Component Analysis, an extension of PCA, achieves non-linear dimensionality reduction using kernels. It supports both transform and inverse_transform.

The Scikit-learn ML library provides sklearn.decomposition.KernelPCA module.

Example

The below example will use sklearn.decomposition.KernelPCA module on Sklearn digit dataset. We are using sigmoid kernel.

from sklearn.datasets import load_digits
from sklearn.decomposition import KernelPCA
X, _ = load_digits(return_X_y = True)
transformer = KernelPCA(n_components = 10, kernel = 'sigmoid')
X_transformed = transformer.fit_transform(X)
X_transformed.shape

Output

(1797, 10)

PCA using randomized SVD

Principal Component Analysis (PCA) using randomized SVD is used to project data to a lower-dimensional space preserving most of the variance by dropping the singular vector of components associated with lower singular values. Here, the sklearn.decomposition.PCA module with the optional parameter svd_solver=’randomized’ is going to be very useful.

Example

The below example will use sklearn.decomposition.PCA module with the optional parameter svd_solver=’randomized’ to find best 7 Principal components from Pima Indians Diabetes dataset.

from pandas import read_csv
from sklearn.decomposition import PCA
path = r'C:\Users\Leekha\Desktop\pima-indians-diabetes.csv'
names = ['preg', 'plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', 'class']
dataframe = read_csv(path, names = names)
array = dataframe.values
X = array[:,0:8]
Y = array[:,8]
pca = PCA(n_components = 7,svd_solver = 'randomized')
fit = pca.fit(X)
print(("Explained Variance: %s") % (fit.explained_variance_ratio_))
print(fit.components_)

Output

Explained Variance: [8.88546635e-01 6.15907837e-02 2.57901189e-02 1.30861374e-027.44093864e-03 3.02614919e-03 5.12444875e-04]
[
   [-2.02176587e-03 9.78115765e-02 1.60930503e-02 6.07566861e-029.93110844e-01 1.40108085e-02 5.37167919e-04 -3.56474430e-03]
   [-2.26488861e-02 -9.72210040e-01 -1.41909330e-01 5.78614699e-029.46266913e-02 -4.69729766e-02 -8.16804621e-04 -1.40168181e-01]
   [-2.24649003e-02 1.43428710e-01 -9.22467192e-01 -3.07013055e-012.09773019e-02 -1.32444542e-01 -6.39983017e-04 -1.25454310e-01]
   [-4.90459604e-02 1.19830016e-01 -2.62742788e-01 8.84369380e-01-6.55503615e-02 1.92801728e-01 2.69908637e-03 -3.01024330e-01]
   [ 1.51612874e-01 -8.79407680e-02 -2.32165009e-01 2.59973487e-01-1.72312241e-04 2.14744823e-02 1.64080684e-03 9.20504903e-01]
   [-5.04730888e-03 5.07391813e-02 7.56365525e-02 2.21363068e-01-6.13326472e-03 -9.70776708e-01 -2.02903702e-03 -1.51133239e-02]
   [ 9.86672995e-01 8.83426114e-04 -1.22975947e-03 -3.76444746e-041.42307394e-03 -2.73046214e-03 -6.34402965e-03 -1.62555343e-01]
]