A Visual Analytics Framework for Spatiotemporal Trade Network Analysis

Abstract

Economic globalization is increasing connectedness among regions of the world, creating complex interdependencies within various supply chains. Recent studies have indicated that changes and disruptions within such networks can serve as indicators for increased risks of violence and armed conflicts. This is especially true of countries that may not be able to compete for scarce commodities during supply shocks. Thus, network-induced vulnerability to supply disruption is typically exported from wealthier populations to disadvantaged populations. As such, researchers and stakeholders concerned with supply chains, political science, environmental studies, etc. need tools to explore the complex dynamics within global trade networks and how the structure of these networks relates to regional instability. However, the multivariate, spatiotemporal nature of the network structure creates a bottleneck in the extraction and analysis of correlations and anomalies for exploratory data analysis and hypothesis generation. Working closely with experts in political science and sustainability, we have developed a highly coordinated, multi-view framework that utilizes anomaly detection, network analytics, and spatiotemporal visualization methods for exploring the relationship between global trade networks and regional instability. Requirements for analysis and initial research questions to be investigated are elicited from domain experts, and a variety of visual encoding techniques for rapid assessment of analysis and correlations between trade goods, network patterns, and time series signatures are explored. We demonstrate the application of our framework through case studies focusing on armed conflicts in Africa, regional instability measures, and their relationship to international global trade.

NEVA: Visual Analytics to Identify Fraudulent Networks

“…Wang et al . [WLS*18] proposed a system of multiple coordinated views that support anomaly detection for global trade network analysis. The system takes localization and events (i.e.…”

Assessing 2D and 3D Heatmaps for Comparative Analysis

“…They compared multiple of these heatmaps with each other by picking out a location of interest in a heatmap and reference the same location in a second heatmap. Various other examples for comparative tasks on heatmaps exist that compare values at a specific location in multiple heatmaps [42,12,58]. To avoid the unfair comparison between the visual variables color and height, we abstained from asking participants to extract exact values from heatmaps and created a task in which participants should estimate the distances between pairs of locations and compare the relative difference of distances.…”

OoDAnalyzer: Interactive Analysis of Out-of-Distribution Samples

“…It aims to identify samples that are inconsistent with the remainder of that set of data [9], [10]. Existing efforts can be classified into two categories [11]: sequence-based methods [12], [13], [14], [15], [16] and point-based methods [17], [18], [19], [20]. Our work is relevant to point-based methods, so we focus on reviewing this category.…”