Reasoning about socio-economic data: a visual analytics approach to Bayesian network

ABSTRACT

The visual analytics approach has become an important tool for analytical reasoning in the fields of information visualization, scientific visualization and cartography. Much research effort has been made to embed statistical methods in the interactive environment of visual analytics where analysts are supported by data visualization tools for the understanding and exploration of complex datasets. This has stimulated further demands on a more collaborative and co-creative human-computer reasoning. In this study, we have developed a model-based visual analytics prototype in which a probabilistic graphical model, namely Bayesian Network, is integrated within a geospatial visualization tool. Such a visual analytics environment aims to empower the joint human-computer reasoning under conditions of uncertainty, where domain experts may interactively assess multiple geospatial datasets. The proposed approach is demonstrated by means of a scenario with heterogeneous socio-economic data in Munich, Germany. By developing a Bayesian Network-enabled visual analytics, we establish a novel system that supports prior knowledge awareness and user involvement in socio-economic data discovery using interactive visualization. The proposed system uses data to uncover structure of the relationships and dependencies among demographic, social and economic factors.

RÉSUMÉ

L'approche basée sur l'analyse visuelle est devenue un outil important pour les raisonnements analytiques dans les domaines de la visualisation de l'information, de la visualisation scientifique et de la cartographie. Beaucoup d'efforts de recherche ont permis d'intégrer des mthodes statistiques dans les environnements interactifs des outils de visualisation analytique o les analystes sont aidés grâce à des outils de visualisation de données permettant de comprendre et d'explorer des ensembles de données complexes. Ces ajouts ont stimulé de nouvelles demandes pour des raisonnements homme-machine plus collaboratifs et co-créatifs. Dans cette étude, nous avons développé un prototype de visualisation analytique à base de modèle, dans lequel un modèle graphique probabilistique, appelé le réseau Bayésien, est intégré à l'intérieur de l'outil de visualisation géospatial. Cet environnement visuel analytique a pour but de renforcer le raisonnement homme-machine dans les conditions d'incertitude où des experts peuvent évaluer interactivement de multiples jeux de données géospatiales. L'approche proposée est validée à l'aide de scénarios sur des données socio-économiques de la ville de Munich en Allemagne. En développant un outil d'analyse visuelle de réseau Bayésien nous réalisons un nouveau système qui intègre des connaissances préalables et la participation d”utilisateurs dans la découverte de données socio-économiques à l'aide de visualisation interactive. Le système proposé utilise des données pour découvrir les structures des relations et des dépendances entre facteurs démographiques, sociaux et économiques.

KEYWORDS:

• Visual analytics
• Bayesian reasoning
• statistical visualization
• socio-economic analysis

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

E. Chuprikova (* 1987) conducts research in the field of cartography and data visualization. In 2010 she received a Diploma degree in Cartography from St. Petersburg State University. She graduated from the International Master Program in Cartography taught by the four cooperating universities: TU Munich, TU Vienna, TU Dresden and the University of Twente. While working for the Institute for Earth Observation, EURAC Research (Bolzano, Italy), she conducted research on the Earth Observation metadata and WebGIS development for rooftop solar potential visualization. Since 2015 she is working as a research associate at the Chair of Cartography, Technical University of Munich, Germany. Ekaterina’s Ph.D. defense took place in January 2019 on the topic ‘Visualizing Uncertainty in Reasoning – A Bayesian Network-enabled Visual Analytics Approach for Geospatial Data’.

Prof. L. Meng (*1963) conducts research in the field of geodesy and geoinformation. Her recent research focus includes geospatial cognition, geodata integration and generalization, non-photorealistic visualization, visual analytics, map-based multimodal navigation services, open portal for geospatial events. Prof. Meng earned her M.Sc. in cartography and geodetic engineering in 1985 in China and her Ph.D. in geodetic engineering at the University of Hannover in 1993. She earned her university teaching qualification in 1998 at the Royal Institute of Technology, Sweden, and was appointed to the Chair of Cartography at TUM the same year. From 2009 to 2012, Prof. Meng was Senator of the Helmholtz Association for the research field Earth and Environment. From 2008 to 2014, she served as Senior Vice President of TUM for international alliances and alumni. She has been a member of the German National Academy of Sciences since 2011 and of the Bavarian Academy of Sciences since 2013.

ORCID

E. Chuprikova http://orcid.org/0000-0002-8981-0442

L. Meng http://orcid.org/0000-0001-8787-3418

Additional information

Funding

This work is supported by Jiangsu Industry Technology Research Institute, Changshu Fengfan Power Equipment Co., Ltd., International Graduate School of Science and Engineering (IGSSE), and the Technical University of Munich (TUM).