E-Embed: A time series visualization framework based on earth mover's distance. (October 2018)

• Record Type: Journal Article
• Authors: Chen, Bingkun; Zhou, Hong; Chen, Xiaojun
• Is Part Of: Journal of visual languages & computing. Volume 48(2018)
• Page Start: 110
• Page End: 122
• Publication Date: 2018-10
• Subjects: Time series -- Visualization -- EMD -- Dimensionality reduction -- Similarity measures; Visual programming languages (Computer science) -- Periodicals; Visual programming (Computer science) -- Periodicals; Programming languages (Electronic computers) -- Semantics -- Periodicals; Langages de programmation visuelle -- Périodiques; Programmation visuelle -- Périodiques; Langages de programmation -- Sémantique -- Périodiques; Programming languages (Electronic computers) -- Semantics; Visual programming (Computer science); Visual programming languages (Computer science); Periodicals; Electronic journals; 005
• DOI: 10.1016/j.jvlc.2018.08.002
• Languages: English
• ISSNs: 1045-926X

Abstract

Abstract: Time series analysis is an important topic in machine learning and a suitable visualization method can be used to facilitate the work of data mining. In this paper, we propose E-Embed: a novel framework to visualize time series data by projecting them into a low-dimensional space while capturing the underlying data structure. In the E-Embed framework, we use discrete distributions to model time series and measure the distances between them by using earth mover's distance (EMD). After the distances between time series are calculated, we can visualize the data by dimensionality reduction algorithms. To combine different dimensionality reduction methods (such as Isomap) that depend on K-nearest neighbor (KNN) graph effectively, we propose an algorithm for constructing a KNN graph based on the earth mover's distance. We evaluate our visualization framework on both univariate time series data and multivariate time series data. Experimental results demonstrate that E-Embed can provide high quality visualization with low computational cost.