Abstract—Today, wide important advances in clustering time series have been obtained in the field of data mining. A large part of these successes are due to the novel achieves in dimensionality reduction and distance measurements of time series data. However, addressing the problem of time series clustering through conventional approach has not solved the issue completely, especially when the class label of time series are vague. In this paper, a two-level fuzzy clustering strategy is employed in order to achieve the objective. In the first level, upon dimensionality reduction by a symbolic representation, time series data are clustered in a high-level phase using the longest common subsequence as similarity measurement. Then, by utilizing an efficient method, prototypes are made based on constructed clusters and passed to the next level to be reused as initial centroids. Afterwards, a fuzzy clustering approach is utilized to justify the clusters precisely. We will present the benefits of the proposed system by implementing a real application: Credit card Transactions Clustering.
Index Terms—Clustering, time series, fuzzy C-mean, longest common subsequence.
Saeed Aghabozorgi is with the Department of Information System, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia (e-mail: firstname.lastname@example.org).
Teh Ying Wah is with the Department of Information System, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia (e-mail: email@example.com).
Cite: Saeed Aghabozorgi and Teh Ying Wah, "Effective Clustering of Time-Series Data Using FCM," International Journal of Machine Learning and Computing vol.4, no. 2, pp. 170-176, 2014.