pysdoclust-stream

incremental stream clustering algorithm based on SDO

Usage in python

from dSalmon import clustering
import numpy as np

size = 500
X = np.random.rand(size, 3)

k = 50 # Model size
T = 150 # Time Horizon
classifier = clustering.SDOcluststream(k=k, T=T)

# Initialize an array to store the labels
all_labels = []
block_size = 25
# Iterate through the data in blocks
for i in range(0, X.shape[0], block_size):
    chunk = X[i:i + block_size, :]
    labels = classifier.fit_predict(chunk)
    all_labels.append(labels)
final_labels = np.concatenate(all_labels)

# print model
obs = classifier.get_observers()
for o in obs:
      print(o)

# print labels
for l in all_labels:
      print(l)

Testing

Tests for this project are located in the tests/temporal-silhouette/ folder. The test framework is mostly adapted from py-temporal-silhouette.

Rebuilding

When adding new algorithms or modifying the interface, the SWIG wrappers have to be rebuilt. To this end, SWIG has to be installed and a pip package can be created and installed using

make && pip3 install dSalmon.tar.xz

Aknowledgments

We would like to thank the developers of the dSalmon project for providing the framework and algorithms, in particular the MTree implementation, that were instrumental in the development of this project.