Article Preview
Top1. Introduction
One of the most important problems in unsupervised learning is clustering. In the clustering process, the summary generation and the exploratory data are the advantages of time series data (Keshavarzi, et al., 2020). This clustering process is also considered to be the pre-processing step in the part of complex systems or in other time-series mining task (Majumdar & Laha, 2020). In static data, the problems that occurred in the process of clustering is studied (for the past 60 years), and some algorithms are considered as the classic to those problems (Ruiz, et al., 2020; Crnkić, et al., 2020; Johnpaul, et al., 2020). Moreover, the clustering method in time series sequences and the clustering of conventional data points are not equal. An unsupervised clustering process includes the two major aspects, such as the appropriate selection of distance measures and the right number of clusters. The initial step in the clustering process is formed by selecting the appropriate distance measures (Du, et al., 2020; Putri, et al., 2019). The distance-based clustering takes more time for clustering only when the sequences get increased. The performance of the clustering process is improved to a small extent based on the uses of different dimensionality reduction methods. Apart from the numerical values, the structure of clustering will highlight the development of time series sequences that assist in fast clustering and reduced dimensionality (Pinto, et al., 2019; Bhaskaran, et al., 2020).
Certain dimensionality reduction techniques such as wavelets, Principal Component Analysis (PCA), etc., were used in time series data clustering The clustering aids in understanding the presence of similar time series sequences (Kamalzadeh, et al., 2020; Li & Wei, 2020). The clustering process of time series data is used in numerous applications like intervention analysis, recognizing significant data points, pattern recognition with the time period, classification of new time series sequences, decision making, etc. (Delforge, et al., 2020; Sun, et al., 2020). The time-series data clustering is classified into three processes as feature-based, model-based and shape-based (China, 2016; Indhumathi & Mohana, 2013; Mohana, 2020). Furthermore, the probabilistic framework is provided by the Bayesian clustering (Devagnanam & Elango, 2020) process that calculates the fit among the model and a dataset with the benefits of prior information (Ribeiro & Rios, 2020; Mahmoudi, et al., 2020; Li, et al., 2020). In unlabeled data, clustering (Veeraiah & Krishna, 2018; Brajula & Praveena, 2018) is one of the descriptive data mining tasks in which the same structures are divided into clusters without any information regarding the definition of groups (Li, et al., 2019; Salgado, et al., 2017; Zhou, et al., 2018).