Basics

• About Gephi
• Importing  graph data
• Interface overview
• Graph layouts

Advanced

• Statistics in Gephi
• Filters
• Clustering
• Dynamic networks

Gephi is

• an open source software for exploring and manipulating networks 
• "like Photoshop but for graph data" (gephi.org)
• written in Java and runs on most OS's
• extensible (plugins with additional features)
• ...a bit buggy :) 

Let's get it started!

Do you have Gephi installed?

gephi.org

Intermission: CSV

• Comma-separated values 
• The simplest table format
• Not limited to comma (anything can serve as a separator — ;$: etc)
• Can be opened in Excel/OpenOffice

Let's try it

https://goo.gl/a0IKQk
(CSV file)

Importing

• File (Файл) -> Import spreadsheet
• Important: Edges table (Таблица рёбер)
• Далее -> Готово

Main windows

• Preview (Обработка)
  • layout and all the calculations
• Data laboratory (Лаборатория данных)
  • your original data (can be modified)
• Overview (Просмотр)
  • for pretty pictures :)

Let's make Gephi show labels

• Data Laboratory (Лаборатория данных)
  • Copy data to another column
    (Скопировать значения в другой столбец)
    • 'Id'
      • Copy to: 'Label'
        (Укажите целевой столбец: Label)

Graph drawing 

укладка графа

• is a complex scientific area on its own
• lies at the intersection of computer science and some heavy-weight mathematics
• can emphasize different features of a graph

Layouts in Gephi

Layouts in Gephi

Let's try some layouts!

Layout/укладка

Statistics 

• Window -> Context  (Окно -> Контекст)
  • Number of nodes and edges
• Window -> Statistics (Окно -> Статистика)
  • ​Click 'Run' to calculate metrics

Now let's see who has the highest degree!

Go to Data Laboratory 

Vizualizing statistics

• Appearance
  • Size
    • Attribute
      • Degree

Clustering in Gephi

• Louvain algorithm (Blondel et al, 2008) 
• The value to be optimized is modularity
• modularity measures the density of links inside communities compared to links between communities.

Formal definition

Intuition behind it

Modularity parameter

• >1 to get less clusters
• <1 (e.g. 0,5) to get more clusters 

Let us try and cluster Hamlet now

Bonus: directed Hamlet graph

for those who want to try it:

https://goo.gl/uJkJXc

Title Text