Skip to content

Latest commit

 

History

History
83 lines (71 loc) · 5.15 KB

README.md

File metadata and controls

83 lines (71 loc) · 5.15 KB
Raw

nashnetx

Network analysis is a powerful and increasingly widespread way to analyze relational data, such as social networks. In this tutorial, we will learn the basics of graph theory and how to use NetworkX, a popular open-source Python package. We'll then apply this knowledge to extract insights about the social fabric of Tennessee MeetUp groups. Blog posts based off this work are (will be) available on stkbailey.github.io.

Forkable and editable kernels are now also available on Kaggle.

Repository Contents

  1. 1_Motivation-Approach.ipynb: Introduction to NetworkX, building and plotting basic graphs.
  2. 2_Getting-MeetUp-Data.ipynb: Introduction to NetworkX, building and plotting basic graphs.
  3. 3_Basic-NetworkX.ipynb: Introduction to NetworkX, building and plotting basic graphs.
  4. 4_PyNash-Relationships.ipynb: Analysis of the PyNash MeetUp group and a ranking of it's most popular members.
  5. 5_Nashville-Relationshisp.ipynb: Analysis of the whole Nashville network, including the group structure.
  6. utils.py: Miscellaneous functions, primarily for interfacing with MeetUp's REST API.
  7. data/: Folder containing data downloaded from Meetup.
  8. presentations/: PowerPoint presentations presented with this material.

Learning goals

  1. Define what a "graph" is and why we would use them.
  2. Describe what Degree, Path Length, Clustering and Centrality measure.
  3. Build and describe a graph in NetworkX.
  4. Leverage graph measures to answer a problem.
  5. Plot a simple graph.

Detailed Outline

  1. Pose the networking problem
    • You're new to Nashville. You're ready to start meeting people, making friends, expanding your network. But you don't want to get pigeon-holed into just one group, i.e. the people you might already know / are comfortable with. Maybe you're trying to market yourself / a product. What groups should you join? Who should you talk to in those groups? Wouldn't you rather be invited into a group than just show up at one?
  2. Agenda
    • Netwhat?
    • Making networks
    • Describing networks
  3. Netwhat?
    • Konigberg bridge problem
      • Display picture of bridges, pose problem: can you walk across all bridges once? Have group take a minute to ponder: how is it possible to solve this?
        • Why or why not?
      • Walk through Euler's thinking...
      • GT describes relationships
    • Examples of usefulness
      • Internet
      • Social networks
      • Brain
    • Vocabulary
      • Nodes, edges, communities, hubs
  4. Making networks
    • Build Konigberg bridge problem
      • Construct graph
        • Discuss different graphs
        • Add nodes
        • Add edges
        • Add attributes
      • Discuss data structure: dict-of-dicts-of-dicts
    • Building MeetUp graph
      • Orient to available data: member data, group data, events attended data
      • Build bipartite graph from DataFrame
      • Make group and member graphs from bipartite graph
  5. Describing networks
    • Orient to the members graph - each edge represents a shared group membership
    • Degree: the people with a lot of connections
      • But maybe they just join big groups? Theyre "mainstream"
    • Path Length: on average, how far is this person from some other person?
      • 7 degress of separation
      • Interesting anecodote about mailers to NY
    • Centrality: how are the people who are "connectors"?
    • Creating a "scorecard" for individuals, then picking out individuals w/ the highest centrality
    • Big reveal: who are the most "central" people in PyNash?
  6. Drawing networks
    • Basics: nodes plotted to points, edges drawn between
    • Gets tricky with larger data
    • Variable sizing of nodes, colors of edges
  7. Credits and point to source code

About the Presenter

I am a PhD student at Vanderbilt, and I enjoy speaking. Although I have never given a large tutorial, I have experience teaching adults in classroom settings, as well as giving lectures to mid-size audiences (30-60). A couple years ago, I attended Toastmasters and got to practice on my fundamental public speaking skills.

I have some expertise in the subject matter from my research, which uses graph theory to analyze brain networks. I have also completed an online course - Applied Social Network Analysis in Python - which has given me some thoughts on how to present the material cogently.

I'll be presenting a brief talk on the proposed material at PyNash on 11/16, which will provide a good opportunity to vet and refine the basic idea. For the tutorial, I would expand the topic a bit and set up a Jupyter Notebook server so that people can follow along and also practice on their own (although I haven't done this before). If I can get that to work, then my actual talk might be a mix of presentation + pseudo-live coding (indexing the graph in different ways, showing how different tweaks can change network properties, etc.).

Data and code would be made available on Github of course. I would also be open to making this a two-hour tutorial and adding in another dataset (brain data) and/or a problem that leverages using graph theory to answer a machine learning problem.

  • font: source code pro, consulus
  • submit to PyTN tutorial