Quickstart¶

Installation¶

The easiest and fastest way to use this package is by downloading it and importing it in your own code (see examples below).

$ git clone https://github.com/albanda/tclust.git

Soon, you will be able to install it through pip (NOTE this option is not available yet):

$ pip install tclust

Getting started¶

As in any sklearn estimator, a typical tclust analysis consists of 3 steps:

Instantiate the class.
Fit the TClust estimator to your data.
You can choose to visualize the output and evaluate your results.

from tclust import TClust
model = TClust(k=3)
model.fit(X)
print(model.labels_)

The examples below show some basic analysis that can be carried out with tclust.

Example 1: Running mixture model¶

In examples/x.csv we provide the dataset for this example. You can follow these instructions to test tclust:

from tclust import TClust
import pandas as pd  # to load the data, stored in a csv file
import matplotlib.pyplot as plt
from sklearn.metrics import confusion_matrix

x = pd.read_csv('x.csv', header=None).values
y = [0] * nsamp + [1] * nsamp + [2] * nsamp + [3] * nsamp

clustering = TClust(k=3, alpha=0.25, n_inits=200, ksteps=40, equal_weights=False, restr_cov_value='deter',
                    maxfact_e=1e10, maxfact_d=10, m=1.1, zero_tol=1e-16, trace=0, opt='mixture', sol_ini=None,
                    tk=False)
clustering.fit(x)
print(set(clustering.best_iter.labels_))
print(confusion_matrix(label, clustering.best_iter.labels_))
plt.subplot(1, 2, 1)
plt.scatter(x[:, 0], x[:, 1], c=y)  # color defined by the original label
plt.xlabel("Feature 1")
plt.xlabel("Feature 2")
plt.title('Ground truth')
plt.scatter(x[:, 0], x[:, 1], c=clustering.best_iter.labels_)  # color defined by tclust
plt.xlabel("Feature 1")
plt.xlabel("Feature 2")
plt.title('Ground truth')
plt.show()

You can look at the plot of the results in the examples/tclust.ipynb notebook.

Example 2: Comparison with existing clustering methods in scikit-learn¶

The figure below shows a comparison of the vanilla version of tclust (i.e., using default parameters) with existing clustering models in scikit-learn. The code to generate the plot is in examples/plot_cluster_comparison_sklearn.ipynb.