Networkt - Temporal Network Analysis

Networkt is a project that aims to do temporal network analysis on digital networks. Temporal network analysis is concerned with the content and time of messages as they traverse the network. This type of analysis allows answering questions about how information flows through networks, and which are the key individuals responsible for diffusing information (particularly innovation). Conversely, traditional network analysis attempts to answer similar questions, but instead by analysis strongly concerned with the relationships between the nodes in a network.

What's the context of this project?

The context of this project is Transnational Entrepreneurship. There is a business theory that states there are a set of individuals named Transnational Entrepreneurs who are responsible for the diffusion of information and innovation across national borders. An example of what a Transnational Entrepreneur may be is an individual with two or more distinct networks in two or more countries. That is, imagine an individual who has a set of friends in country A and a set of friends in country B. It is said that this individual is responsible for spreading ideas and innovations between these countries.

What's the question you're trying to answer?

A simplified explanation for what we are trying to prove is the following: Is the "diversity" of an entrepreneur's network a moderator for how frequently they diffusion innovation and information across networks and borders.

To find out more please check out the /report directory where you can find out more information about this research project and how it was conducted.

How does your software work? What's your general approach?

Our software is written in Python and designed to be easily executable so that you can use it for tests and information gathering of your own. Below, briefly described are the tools we use, and the process of of our software / analysis.

Information about our setup

• We use the twitter API to gather all of our data about a network
• We use Neo4j for our network persistence to disk
• We use scikit learn for all of our network analysis and content clustering

Information about our process

1. Select a startup incubator/workspace/hotspot in a city we are interested in studying, find their twitter username.
2. Collect a set of followers following that startup incubator, living in the same city.
3. For each collected follower (F) following the startup incubator/workspace/hotspot, collect a sample of their network.
4. Check the sample network of (F) to see whether it qualifies them as a Transnational (the details qualifying a network can be found in the report).
5. If the follower (F) qualifies, collect a larger version of their network for analysis. This entails getting a larger set of their friends and followers.
6. Collect the last 200 tweets of every individual in the follower's (F) egocentric network.
7. Run content clustering on tweets within (F)'s network to determine how similar two tweets are. If they are similar within some threshold (defined in the report), label them as being about the same topic.
8. Finally, determine how many instances occur of a tweet traversing the network in such a way that the information traveled through our Transnational entrepreneur. That is, how often is the Transnational entrepreneur (F) responsible for spreading information between the distinct networks (networks delineated by country) they are part of. To do this, we see if a tweet was tweeted by the friend of our Transnational entrepreneur, then by the Transnational entrepreneur, then by a follower of the Transnational entrepreneur. It is important that the friend and follower tweeting are from two distinct networks (this helps ensure to some degree that the Transnational entrepreneur's follower received the information from the Transnational themselves. If the above was confusing, the diagram below may help clarify the interaction we are looking for.

+---------------+      +---------------+     +-----------------+
| Person A      |      | Transnational |     | Person B        |
| Country A     |      | Entrepreneur  |     | Country B       |
| Tweet Topic 1 +------> Tweet Topic 1 +-----> Tweet Topic 1   |
|               |      | Network A, B  |     |                 |
|               |      |               |     |                 |
+---------------+      +---------------+     +-----------------+

How can I run your software?

Please check the readme in the /source folder for all details, including installation, execution, and how you can recreate this project yourself.

What's the origin of the name?

The name derives from 'network' + 'time (t)' - hence, networkt.