IMT Atlantique - Bretagne Pays de la Loire

#
      # Community dynamics analysis
      ## Lab-STICC (CNRS, UMR 6285)
      ### Cécile Bothorel, Laurent Brisson

---

layout:true

.footer[
      # Community dynamics analysis
      ## [DECIDE](https://labsticc.fr/en/teams/decide) TEAM, [Lab-STICC](https://labsticc.fr/en) Laboratory (CRNS, UMR 6285)
      ## [Cécile Bothorel](https://www.imt-atlantique.fr/fr/personne/cecile-bothorel), [Laurent Brisson](http://www.laurent-brisson.fr/)
      .logo-labsticc[]
      ]

---
      name: grid-layout
      .header[
      # Community Definition
      ]

The term community is used here in the sense used in the field of **social network analysis**.

## Definition

It is a grouping of individuals created by applying a clustering algorithm that maximizes interactions within groups and minimizes interactions between groups.

---

<center><span class="green">Business</span> Tasks and <span class="blue">Data Scientist</span> Tasks</center>

<div class="container-5x2">
        <div class="row1col1 flex">
          <i class="fa-duotone fa-square-1 green" fa-3x></i>
          Discussion Threads
        </div>
        <div class="row1col2 flex">
          <i class="fa-duotone fa-square-2 blue" fa-3x></i>
          Interactions Graphs
        </div>
        <div class="row1col3 flex">
          <i class="fa-duotone fa-square-3 blue" fa-3x></i>
          Evolving Communities
        </div>
        <div class="row1col4 flex">
          <i class="fa-duotone fa-square-4 green" fa-3x></i><i class="fa-duotone fa-square-4 blue" fa-3x></i>
          Metrics Computation
        </div>
        <div class="row1col5 flex">
          <i class="fa-duotone fa-square-5 green" fa-3x></i>
          Business Interpretation
        </div>
        <div class="row2col1 flex">
          <img src="images/threads.png" style="height:300px;">
        </div>
        <div class="row2col2 flex">
          <img src="images/networking-infographic.png" style="width:100px;">
        </div>
        <div class="row2col3 flex">
          <img src="images/evolving-communities.png" style="width:100px;">
        </div>
        <div class="row2col4 flex">
          <img src="images/kpi_dark.png" style="width:100px;">
        </div>
        <div class="row2col5 flex">
          <img src="images/analyst.png" style="width:100px;">
        </div>
      </div>

---

## Differents kinds of metrics

<div class="container-2x3">
        <div class="row1col1 flex" style='font-weight: bold;'>
          <i class="fa-duotone fa-messages-question green" fa-3x></i>
          Use metrics
        </div>
        <div class="row2col1 flex" style='font-weight: bold;'>
          <i class="fa-duotone fa-hashtag green" fa-3x></i>
          Context-dependant metrics
        </div>
        <div class="row3col1 flex" style='font-weight: bold;'>
          <i class="fa-duotone fa-chart-network blue" fa-3x></i>
          Topological metrics
        </div>
        <div class="row1col2 flex" style="text-align:left;">
          Number of members, messages<br>
          Answer time, duration of threads
        </div>
        <div class="row2col2 flex">
          Toxicity levels<br>
          Topics
        </div>
        <div class="row3col2 flex">
          Static: hub dominance, clustering coefficiant<br>
          Dynamic: events, communities evolutions
        </div>
      </div>

## Different levels of analysis

- static communities (over a given period)
      - evolving communities (over time)
      - platform

---

.left-column50[
      ## Community structures
      ### (On one period)
      ![Community structures](images/community_internal_structure_dark.png)
      <div style="font-size: x-small; line-height: 1.5;">Vinh-Loc Dao. Characterizing community detection algorithms and detected modules in large-scale complex networks. Data Structures and Algorithms. Ecole nationale supérieure Mines-Télécom Atlantique, 2018. <a href="https://tel.archives-ouvertes.fr/tel-02121358">https://tel.archives-ouvertes.fr/tel-02121358</a></div>
      ]

.right-column50[
      ## Community evolution forms
      ### (Between two periods)
      ![Community evolution forms](images/evolutions_dark.png)
      <div style="font-size: x-small; line-height: 1.5;">Palla, G., Barabási, AL. & Vicsek, T. Quantifying social group evolution. Nature 446, 664–667, 2007. <a href="https://doi.org/10.1038/nature05670">https://doi.org/10.1038/nature05670</a></div>
      ]

---

## Backer profiles on the Ulule crowdfunding platform

Analysis of the static structure of the platform's community to detect different profiles of backers: sponsors, pre-cursors, experts, followers, etc.

<div style="font-size: x-small; line-height: 1.5;">Inna Lyubareva, Laurent Brisson, Cécile Bothorel, Romain Billot. Une plateforme de crowdfunding et son réseau social : L'exemple Ulule. Revue Française de Gestion, Lavoisier, 2020, Les mutations de l’accompagnement entrepreneurial, 1 (286), pp.135--151. <a href="https://dx.doi.org/10.3166/rfg.2019.00402">https://dx.doi.org/10.3166/rfg.2019.00402</a></div>

## Learner profiles on the squily e-learning platform

Detection of student behavioral profiles and analysis of posture changes to understand community dynamics in peer review.

<div style="font-size: x-small; line-height: 1.5;">Raphaël Charbey, Laurent Brisson, Cécile Bothorel, Philippe Ruffieux, Serge Garlatti, et al.. Roles in social interactions: graphlets in temporal networks applied to learning analytics. COMPLEX NETWORKS 2019 : 8th International Conference on Complex Networks and their Applications, Dec 2019, Lisbon, Portugal. <a href="https://dx.doi.org/10.1007/978-3-030-36683-4_41">https://dx.doi.org/10.1007/978-3-030-36683-4_41</a></div>

## Detection of echo chambers on YouTube media channels

Dynamic analysis of community structures to detect inward-looking communities by excluding members with differing opinions.

<div style="font-size: x-small; line-height: 1.5;">Project funded by the French National Research Agency. <a href="http://www.anr-pil.org/">http://www.anr-pil.org/</a></div>

MathJax.Hub.Configured();