Semi-supervised clustering with COBRAS

Library for semi-supervised clustering using pairwise constraints.

COBRAS supports three modes for constraint elicitation:

1. With labeled data. in this case the pairwise relations are derived from the labels. This is mainly used to compare COBRAS experimentally to competitors.
2. With interaction through the commandline. In this case the user is queried about the pairwise relations, and can answer with yes (y) and no (n) through the commandline. The indices that are shown in the queries are the row indices in the specified data matrix (starting from zero).
3. With interaction through a visual user interface. If you use COBRAS-TS, the instantiation of COBRAS that is tailored to time series clustering, you can use an interactive web application that visualizes the data, queries, and intermediate clustering results. A demo can be found at https://dtai.cs.kuleuven.be/software/cobras/

Installation

This package is available on PyPi:

$ pip install cobras_ts

The following dependencies are automatically installed: dtaidistance, kshape, numpy, scikit-learn.

In case you want to use the interactive GUI, install cobras_ts using the following command to automatically install additional dependencies (bokeh, datashader, and cloudpickle):

$ pip install --find-links https://dtai.cs.kuleuven.be/software/cobras/datashader.html pip cobras_ts[gui]

Usage

COBRAS from the command line

The COBRAS algorithm can easily be run from the command line. A cobras_ts script will be installed by pip:

$ cobras_ts --format=csv --labelcol=0 /path/to/UCR_TS_Archive_2015/ECG200/ECG200_TEST

This script is also available in the repository as cobras_ts_cli.py.

COBRAS as a Python package

Examples can also be found in the examples subdirectory.

Running COBRAS_kmeans:

import numpy as np
from sklearn import metrics

from cobras_ts.cobras_kmeans import COBRAS_kmeans
from cobras_ts.labelquerier import LabelQuerier

budget = 100

data = np.loadtxt('/home/toon/data/iris.data', delimiter=',')
X = data[:,1:]
labels = data[:,0]

clusterer = COBRAS_kmeans(X, LabelQuerier(labels), budget)
clusterings, runtimes, ml, cl = clusterer.cluster()

final_clustering = clusterings[-1].construct_cluster_labeling()
print(metrics.adjusted_rand_score(final_clustering,labels))

Running COBRAS_kShape:

import os

import numpy as np
from sklearn import metrics

from cobras_ts.cobras_kshape import COBRAS_kShape
from cobras_ts.labelquerier import LabelQuerier

ucr_path = '/home/toon/Downloads/UCR_TS_Archive_2015'
dataset = 'ECG200'
budget = 100

data = np.loadtxt(os.path.join(ucr_path,dataset,dataset + '_TEST'), delimiter=',')
series = data[:,1:]
labels = data[:,0]

clusterer = COBRAS_kShape(series, LabelQuerier(labels), budget)
clusterings, runtimes, ml, cl = clusterer.cluster()

final_clustering = clusterings[-1].construct_cluster_labeling()
print(metrics.adjusted_rand_score(final_clustering,labels))

Running COBRAS_DTW:

This uses the dtaidistance package to compute the DTW distance matrix. Note that constructing this matrix is typically the most time consuming step, and significant speedups can be achieved by using the C implementation in the dtaidistance package.

import os

import numpy as np
from dtaidistance import dtw
from sklearn import metrics

from cobras_ts.cobras_dtw import COBRAS_DTW
from cobras_ts.labelquerier import LabelQuerier

ucr_path = '/home/toon/Downloads/UCR_TS_Archive_2015'
dataset = 'ECG200'
budget = 100
alpha = 0.5
window = 10

data = np.loadtxt(os.path.join(ucr_path,dataset,dataset + '_TEST'), delimiter=',')
series = data[:,1:]
labels = data[:,0]


dists = dtw.distance_matrix(series, window=int(0.01 * window * series.shape[1]))
dists[dists == np.inf] = 0
dists = dists + dists.T - np.diag(np.diag(dists))
affinities = np.exp(-dists * alpha)

clusterer = COBRAS_DTW(affinities, LabelQuerier(labels), budget)
clusterings, runtimes, ml, cl = clusterer.cluster()

final_clustering = clusterings[-1].construct_cluster_labeling()
print(metrics.adjusted_rand_score(final_clustering,labels))

Dependencies

This package uses Python3, numpy, scikit-learn, kshape and dtaidistance.

Contact

Toon Van Craenendonck at toon.vancraenendonck@cs.kuleuven.be

License

COBRAS code for semi-supervised time series clustering.

Copyright 2018 KU Leuven, DTAI Research Group

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.