Time Series Projection to Highlight Trends and Outliers

Abstract

A B C Figure 1: ViCCEx -Visual Chemical Contamination Explorer (A) The t-SNE projection enables to identify patterns, trends, and outliers in multivariate time series sensor readings (B) Sampling strategy view to investigating the chemicals measurements taken at each location (C) Metadata panel with filter options and further extracted statistics for each chemical for each sensor location. ABSTRACTThe goal of the VAST Challenge 2018 Mini Challenge 2 (MC 2) was to unveil the possible causes and effects of environmental pollution in the Boonsong Lekagul Wildlife Preserve. We propose the ViCCEx (Visual Chemical Contamination Explorer) system that enables to interactively explore the sparse multivariate river network sensor reading dataset to identify characteristics, trends, and outliers of the different sensor reading locations over time. The ViCCEx system uses a t-SNE projection to display an overview visualization, a sampling strategy view to highlight the overall sampling strategies of different chemical measurements at each sensor location, and various extracted statistics to highlight the evolution of chemical measurements. The three views are connected via linking and brushing, which enables to explore and identify possible pollution causes and effects in the preserve.

ATOVis – A visualisation tool for the detection of financial fraud

“…Its application can be seen in works from diverse domains of application. [23][24][25][26][27][28] From the literature, the timeline representation can be subdivided into two different approaches: faceted and multiscale. 21 The faceted method can be defined as multiple timelines aligned spatially, each one representing the same time interval.…”

“…21 The faceted method can be defined as multiple timelines aligned spatially, each one representing the same time interval. [29][30][31] To represent data variables in faceted timelines, different approaches can be used: bars 24,27,[32][33][34] ; lines 23,26 ; heatmaps 35 ; glyphs 28,36 ; and image and/or text. 15,29,37,38 The multiscale technique can be defined as multiple timelines with different time granularities.…”

ProjectionPathExplorer: Exploring Visual Patterns in Projected Decision-Making Paths

“…In ModelSpace, Brown et al [8] visualize user interaction data as projected trajectories, aiming to reveal patterns in the users' intentions. Cakmak et al used a t-SNE visualization for trend and outlier detection in projected time series [9]. Zhu and Chen created dimensionality-reduced trajectories for basketball games, making use of landmark MDS [63].…”