Add reference on k-means initialization

kgryte · kgryte · commit a8c5e7f5bd4d · 2018-07-13T21:43:23.000+01:00
diff --git a/docs/references/bib.bib b/docs/references/bib.bib
@@ -1975,3 +1975,24 @@ @inproceedings{arthur:2007a
 	url = {http://dl.acm.org/citation.cfm?id=1283383.1283494},
 	year = {2007},
 }
+
+@article{celebi:2013a,
+	abstract = {K-means is undoubtedly the most widely used partitional clustering algorithm. Unfortunately, due to its gradient descent nature, this algorithm is highly sensitive to the initial placement of the cluster centers. Numerous initialization methods have been proposed to address this problem. In this paper, we first present an overview of these methods with an emphasis on their computational efficiency. We then compare eight commonly used linear time complexity initialization methods on a large and diverse collection of data sets using various performance criteria. Finally, we analyze the experimental results using non-parametric statistical tests and provide recommendations for practitioners. We demonstrate that popular initialization methods often perform poorly and that there are in fact strong alternatives to these methods.},
+	acmid = {2369887},
+	address = {Tarrytown, NY, USA},
+	author = {M. Emre Celebi and Hassan A. Kingravi and Patricio A.Vela},
+	doi = {10.1016/j.eswa.2012.07.021},
+	issn = {0957-4174},
+	issue_date = {January, 2013},
+	journal = {Expert Systems with Applications: An International Journal},
+	keywords = {cluster center initialization, partitional clustering, sum of squared error criterion, k-means, machine learning, ml, kmeans, clustering},
+	month = {jan},
+	number = {1},
+	numpages = {11},
+	pages = {200--210},
+	publisher = {Pergamon Press, Inc.},
+	title = {{A Comparative Study of Efficient Initialization Methods for the K-means Clustering Algorithm}},
+	url = {http://dx.doi.org/10.1016/j.eswa.2012.07.021},
+	volume = {40},
+	year = {2013},
+}
