Hybrid clustering algorithm for time series data — A literature survey

Abstract

Mining of Time Series data has an impressive growth of interest in today's world. To provide an indication various implementation mechanisms are studied and summarized the different types of problems identified in existing applications. Clustering time series data is a trouble that has applications in an extensive variety of areas and has recently evoked a large amount of research. Time series data may contain large and outliers. In addition, time series data is a one kind of special data set where attributes have a temporal ordering. Therefore clustering of time series data is a good issue in the data mining process. Different techniques and various clustering algorithms have been proposed to assist clustering of time series data sets also different kinds of non-developmentary optimization techniques and for clustering multivariate in some applications, usually they produces poor efficient results due to the dependency on the initial set of values and their poor performance in manipulating multiple objectives. Sometimes Time series data doesn't contain same length and they usually have missing values, the basic measure Euclidean distance and dynamic time warping cannot be applied for such datasets to measure the similarity of objects.