Biclustering

Biclustering is a form of unsupervised machine learning that takes a data matrix and groups both the rows and columns of this matrix to unveil previously unknown patterns. It’s  standard in gene expression, text mining, and other recommendation systems and captures more localized relationships than the general clustering method. Scikit-learn provides spectral co-clustering and diagonal biclustering algorithms, implemented as classes with a fit method, enabling efficient pattern discovery in complex datasets.

Syntax

Here’s a syntax that shows the implementation of biclustering using sklearn:

from sklearn.cluster import SpectralCoclustering, SpectralBiclustering

# For Spectral Co-clustering
model = SpectralCoclustering(n_clusters=number_of_biclusters, random_state=seed)
model.fit(data_matrix)

# For Spectral Bi-clustering
model = SpectralBiclustering(n_clusters=number_of_biclusters, method="log", random_state=seed)
model.fit(data_matrix)

• n_clusters: Number of biclusters to create.
• random_state: Ensures the randomness for reproducible results.
• method(For SpectralBiclustering): Specifies the algorithm variant, e.g., log or bistochastic. The log method applies logarithmic scaling, while bistochastic normalizes rows and columns. The choice of method can affect the results depending on the dataset.

Note: Since Bicluster is not directly available in sklearn, alternative methods for biclustering, such as SpectralBiclustering, can be used.

Example

Here’s an example of implementing biclustering using SpectralBiclustering from sklearn:

import numpy as np
from sklearn.cluster import SpectralBiclustering

# Sample data matrix
data_matrix = np.array([[1, 1, 0, 0],
                        [1, 1, 0, 0],
                        [0, 0, 1, 1],
                        [0, 0, 1, 1]])

# Apply Spectral Biclustering
model = SpectralBiclustering(n_clusters=2, random_state=42)
model.fit(data_matrix)

# Get the bicluster labels for rows and columns
row_labels = model.rows_
column_labels = model.columns_

# Print biclusters
print("Row Biclusters:", row_labels)
print("Column Biclusters:", column_labels)

The above code results in the following output:

Row Biclusters: [[False False  True  True]
 [False False  True  True]
 [ True  True False False]
 [ True  True False False]]
Column Biclusters: [[False False  True  True]
 [ True  True False False]
 [False False  True  True]
 [ True  True False False]]

• In the Row Biclusters, True in a position means that the corresponding row is part of the bicluster.
• Similarly, in the Column Biclusters, True indicates that the corresponding column is part of the bicluster.

Codebyte Example

Here the example demonstrates how to perform Spectral Biclustering on a simple 6x6 binary data matrix using SpectralBiclustering from sklearn:

us
Visit us
code
Hide code
Code
Output
Hide output
Hide output
Loading...
Run