egoSlider: Visual Analysis of Egocentric Network Evolution Yanhong Wu , Naveen Pitipornvivat, Jian Zhao, Sixiao Yang, Guowei Huang, and Huamin Qu
Definition • Ego-network: the relationships between a specific individual, i.e., the ego and people connected to it, i.e., the alters • Analyzing ego-network evolution provides huge insights to many domains: Sociology Anthropology Psylogy http://www.majorfinder.com/majors/anthropology http://www.differencebetween.com/difference-between-sociology-and-vs-social-science/ 1 http://directactioneverywhere.com/theliberationist/2015/3/10/what-does-psychology-really-tell-us-about-animal-welfarism
Challenges • Analyzing ego-network is challenging due to the complex time- varying structures 2 Icons: https://www.iconfinder.com/iconsets/user-pictures
Challenges • Analyzing ego-network is challenging due to the complex time- varying structures • Alters come and leave 3 Icons: https://www.iconfinder.com/iconsets/user-pictures
Challenges • Analyzing ego-network is challenging due to the complex time- varying structures • Alters come and leave • Ties grow stronger and fade away 4 Icons: https://www.iconfinder.com/iconsets/user-pictures
Challenges • Analyzing ego-network is challenging due to the complex time- varying structures • Alters come and leave • Ties grow stronger and fade away • Alter communities merge and split 5 Icons: https://www.iconfinder.com/iconsets/user-pictures
Motivation • Analyzing ego-network is challenging due to the complex time- varying structures • Alters come and leave • Ties grow stronger and fade away • Alter communities merge and split Connection strengths and inter-alter relations are omitted in existing works! 6
Analytical Questions • Macroscopic Level • What are the overall patterns of a large group of people’s ego-networks at each time step? 7
Analytical Questions • Macroscopic Level • What are the overall patterns of a large group of people’s ego-networks at each time step? • Mesoscopic Level • What are the general similarities between multiple people’s ego-networks along time? 7
Analytical Questions • Macroscopic Level • What are the overall patterns of a large group of people’s ego-networks at each time step? • Mesoscopic Level • What are the general similarities between multiple people’s ego-networks along time? • Microscopic Level • How does an ego’s alter number and the tie strengths evolve over time? How are the alters connected? 7
Demo! 8
Solution • egoSlider: an interactive visualization system that enables users to explore, compare, and analyze dynamic ego-network evolutions egoSlider pipeline 9
Data Overview – Macroscopic Level • Goal: obtain a whole picture of all the ego-networks • Approach: MDS (multidimensional scaling) • Metric: 7 ego-network attributes including alter number, network density, average tie strength… • Distance function: Canberra distance 10
Data Overview – Macroscopic Level • Goal: obtain a whole picture of all the ego-networks • Approach: MDS (multidimensional scaling) Four groups where one is slightly larger One giant cluster, a much smaller cluster, and several outliers 11
Summary Timeline View – Mesoscopic Level • Goal: track and compare the statistical feature changes of multiple ego-networks over time • Approach: snapshot glyph and transition glyph 12
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T Alter type 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T Number of 1-degree alters 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T Ego density 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T Number of 2-degree alters 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T T+1 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T T+1 Number of consistent alters 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T T+1 Alter tie strengths become weaker 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T T+1 Alter tie strengths become stronger 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T T+1 Alters keep the same tie strength 13
Summary Timeline View – Mesoscopic Level • Approach: snapshot glyph and transition glyph Time: T T+1 No consistent 1-degree alters 13
Summary Timeline View – Mesoscopic Level • Case Study 14
Summary Timeline View – Mesoscopic Level • Case Study 14
Alter Timeline View – Microscopic Level • Goal: study the detailed behaviors of a particular individual’s ego- network evolution • Approach: timeline-based visualization 15
Alter Timeline View – Microscopic Level • Approach: timeline-based visualization Alter bars 16
Alter Timeline View – Microscopic Level • Approach: timeline-based visualization 2-degree alter volume flow 16
Alter Timeline View – Microscopic Level • Approach: timeline-based visualization Reemerging 1-dgree alter 16
Alter Timeline View – Microscopic Level • Case Study 17
Alter Timeline View – Microscopic Level • Case Study Prof. Kwan-Liu Ma’s filtered Prof. Daniel Cohen-Or’s filtered ego-network between 2005 - 2010 ego-network between 2005 - 2010 17
Alter Timeline View – Microscopic Level • Case Study Prof. Kwan-Liu Ma’s filtered Prof. Daniel Cohen-Or’s filtered ego-network between 2005 - 2010 ego-network between 2005 - 2010 17
Evaluation • Case study – DBLP dataset • 64,892 authors and 52,038 papers from 31 conferences • User study • 15 participants are recruited • Compared with an improved node-link based system (a) Highlighting nodes (b) Highlighting edges (c) Highlighting clusters 18
User Study - Results • 12 tasks in mesoscopic and microscopic level • 6 significantly better in both response accuracy and response time Response accuracy 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 egoSlider baseline Meso-level tasks Micro-level tasks 19
User Study - Results • 12 tasks in mesoscopic and microscopic level • 6 significantly better in both response accuracy and response time Response time (avg.) Time (s) 30.00 25.00 20.00 15.00 10.00 5.00 0.00 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 egoSlider baseline Meso-level tasks Micro-level tasks 20
Future Work • Investigate ego-network sequence similarities • Incorporate modern multivariate visualization techniques • Conduct more realistic case studies and user studies 21
Q&A egoSlider: Visual Analysis of Egocentric Network Evolution Yanhong Wu , Naveen Pitipornvivat, Jian Zhao, Sixiao Yang, Guowei Huang, and Huamin Qu yanhong.wu@ust.hk http://yhwu.me
Recommend
More recommend
