A segmentation technology for multivariate contextual time series

Abstract

A time series is a series of data points indexed in time order, mining multivariate contextual time series (MCTS) should pay more attention to time order. This paper proposes a new method for splitting the MCTS into a number of segments, uses the concept of scenarios and themes to represent MCTS instead of data points and extracts important contextual features to carry out the multidimensional fitting for MCTS.