Comparison of Clustering Algorithms in Text Clustering Tasks

The purpose of this paper is to compare the performance and accuracy of several clustering algorithms in text clustering tasks. The text preprocessing were realized by using the Term Frequency - Inverse Document Frequency in order to obtain weights for each word in each text and then obtain weights for each text. The Cosine Similarity was used as the similarity measure between the texts. The clustering tasks were realized over the PAN dataset and three different algorithms were used: Affinity Propagation, K-Means and Spectral Clustering. This paper presents the results in comparative tables: ID of the task, ground truth clusters and the clusters generated by the algorithms. A table with precision, recall and f-measure scores is presented.