State of the Art in Patterns for Point Cluster Analysis

Abstract

Abstract. Nowadays, an abundance of sensors are used to collect very large datasets containing spatial points which can be mined and analyzed to extract meaningful patterns. In this article, we focus on different techniques used to summarize and visualize 2D point clusters and discuss their relative strengths. This article focuses on patterns which describe the dispersion of data around a central tendency. These techniques are particularly beneficial for detecting outliers and understanding the spatial density of point clusters.

Trajectory Box Plot: a new pattern to summarize movements

“…In this article, a new spatio-temporal pattern called the Trajectory Box Plot (TBP) which relies on classical Box Plot concept is introduced. TBP extends Oriented Spatial Box Plot (OSBP) (Etienne et al 2014) to a cluster of trajectories with the same itineraries. This pattern is a combination of a reference median trajectory and two ordered sets of 3D boxes.…”

“…• The orientation of the point cluster which is linked with the correlation between the x and y coordinates of the points; Patterns such as the Rangefinder Box Plot, Bivariate box plot and the OSBP extend principles of the traditional box plot (Tukey 1977) which is useful for representing varia-7 tion in samples of statistical one dimensional datasets without making assumptions about the underlying statistical distribution. Etienne et al (2014) highlighted the advantages of OSBP.…”

“…One important feature of a visualisation pattern is its ability to show central tendency, dispersion and symmetry simultaneously. The Oriented Spatio Box Plot (OSBP) (Etienne et al 2014) provides a means to visually understand these statistical parameters for point clusters.…”