CLUSTERING AND CLASSIFICATION OF TIME SERIES
Daniel Peña, Ruey S. Tsay · Wiley series in probability and statistics · 2021
This chapter shows how to divide a set of time series into homogeneous groups of series with similar properties and how to classify a time series into one cluster among several possible clusters. It discusses some measures of distance or dissimilarities between time series. The chapter presents four procedures for clustering analysis based on variables. These methods are k-means, k-medoids that can also be used with dissimilarities, projection methods, and mixtures of distributions. When we use features of the series for the classification, then the standard cross-validation can be applied by splitting the sample into training and testing samples. When the distribution of the features is unknown, several nonparametric methods can be used for discrimination. They are nearest neighbors, support vector machines, and density estimation.