This post will concentrate on the problem of time series clustering and classification. So before entering the technical details, what I mean from clustering and classification :
- Clustering : It is drawn from the non-supervised learning methods and consists of assigning a label to the entire time series data such that the objects (time series) in the same class represent the same dynamics over time.
- Classification : It is drawn from the supervised learning class of methods and consists of determining the future states of a time series on the basis of some historical data. It can be done for each time series independently or over multiple time series simultaneously. In this case, I consider that time series are represented as categorical data and can be associated to the evolution of the state over time.