Zaman serisi verileri için K-medyan kümeleme algoritmaları

Clustering is an unsupervised learning method, that groups the unlabeled data for gathering valuable information. Clustering can be applied on various types of data. In this study, we have focused on time series clustering. When the studies about time series clustering are reviewed in the literature, for the time series data, the centers of the formed clusters are selected from the existing time series samples in the clusters. In this study, we have changed that view and have proposed clustering algorithms based on the idea of selecting the cluster centers for each timestamp. With this view, we aim to improve the clustering performance. Based on this idea four different algorithms are suggested that are called as Center Based K-Median Algorithm (CKM), CKM with Haar Wavelet decomposition, CKM with Haar Wavelet Decomposition Without Projection and Search Based CKM with Haar Wavelet Decomposition. In the first algorithm, the raw data is used and the clustering problem is solved by the proposed optimization model. The other three algorithms are also solved by using the proposed optimization model and instead of using raw data, transformed data, which the Haar wavelet decomposition is applied to, is used. The proposed algorithms have been experimented on different data sets and evaluated by using different internal and external indices. Due to the evaluations, successful results are obtained regarding clustering performances of the CKM based algorithms.