Skip to main content

A fast classifier for binary node classification in bipartite graphs

Project description


The SW-transformation is a fast classifier for binary node classification in bipartite graphs (Stankova et al., 2015). Bipartite graphs (or bigraphs), are defined by having two types of nodes such that edges only exist between nodes of the different type (see Fig. 1).


Fig. 1: Bigraph, top node projection and bottom node projection (left), adjacency matrix representation of the bigraph (right) (Stankova et al., 2015, p. 8).

The SW-transformation combines the weighted-vote Relational Neighbor (wvRN) classifier with an aggregation function that sums the weights of the top nodes. The transformation optimally considers for each test instance only the weights of the neighboring top nodes multiplied by the number of training instances in that column which have a positive label (the positive neighbors of the node). The SW-transformation yields very fast run times and allows easy scaling of the method to big data sets of millions of nodes (Stankova et al., 2015).


Requirements: numpy

pip install numpy

To build and install on your local machine, download and unzip the repository and run from there:

python install

Or you can automatically download and install from the PyPI repository:

pip install SW-transformation


The SW transformation can be used to calculate the probability of a node in a bipartite graph to belong to the positive class. Three top node weight functions are included: simple weight assignment, inverse degree and hyperbolic tangent. Users can also directly insert the top node weights, using their own weight function.

SW-transformation(weight_function='tanh', weights=None)


  • weight_function : 'tanh', 'inverse', 'simple' or 'own' (default='tanh'). For 'own', the top node weights should be provided in weights
  • weights : array-like, shape (1, n_top_nodes). Vector containing the top node weights, where n_top_nodes is the number of top nodes. When weight_function is set at 'own', the top node weights should be provided here.


from sklearn.model_selection import GridSearchCV
import numpy as np
import SW

X = load_adjacency_matrix_here #sparse matrix
y = load_node_labels_here

custom = lambda X: np.ones((X.shape[1],)) # define custom weight function

sw = SW.SW_transformation()
parameters = {'weight_function':('tanh','inverse','simple',custom)}
clf = GridSearchCV(sw, parameters, cv=5, scoring = 'roc_auc', verbose=3),y)


  • fit(X, y) - Fit the model according to the given training data.
  • predict_proba(X) - Probability estimates.

X: sparse matrix, shape(n_bottom_nodes, n_top_nodes): the adjacency matrix (see Fig. 1, right)

y: array_like, shape(n_bottom_nodes, 1): the binary class labels (0 for negative and 1 for positive instance)

Authors and contributors

Stiene Praet



This project is licensed under the MIT License - see the file for details


Based on the work of Marija Stankova, David Martens and Foster Provost


Stankova, M., Martens, D., & Provost, F. (2015). Classification over bipartite graphs through projection. (Research paper / University of Antwerp. Faculty of Applied Economics ; 2015-001 D/2015/1169/001). Full text (open access):

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

SW-transformation-0.0.2.tar.gz (4.0 kB view hashes)

Uploaded source

Built Distribution

SW_transformation-0.0.2-py3-none-any.whl (5.5 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page