Social Data and Networks
Date: July 7-8, 2014 | Tuition: $1,800 | Continuing Education Units (CEUs): 1.4
*This course has limited enrollment. Apply early to guarantee your spot.
Application Deadline »
Networks are ubiquitous in the modern era, be they social media networks such as Facebook, transportation networks formed by aerial or ground routes, political networks observed through blogs and opinions, energy dispatch networks formed between end-users and generators, or meta-networks observed in ‘interdependent’ data such as World Wide Web or biological protein interaction. We know a great deal about these networks, as all things online are recorded, cheap sensors of all sorts are providing a wealth of information about operations, and experiments can be performed at a massive scale. Understanding and utilizing such networks can help tremendously in making better societal decisions regarding public policies, business operations, financial market regulations, and utility network operations.
In this course, we provide an in-depth, state-of-the-art analytic view toward making use of these networks a reality. The course will be about concepts and fundamentals accompanied by appropriate case-studies.
Fundamentals: Core concepts, understandings and tools (30%)
Latest Developments: Recent advances and future trends (40%)
Industry Applications: Linking theory and real-world (30%)
Lecture: Delivery of material in a lecture format (80%)
Discussion or Groupwork: Participatory learning (10%)
Labs: Demonstrations, experiments, simulations (10%)
Introductory: Appropriate for a general audience (70%)
Specialized: Assumes experience in practice area or field (25%)
Advanced: In-depth explorations at the graduate level (5%)
The participants of this course will be able to:
- Utilize social data to make operational decisions for areas such as advertising and marketing, policy making, customer service, etc.
- Develop an effective summary of social data in the form of meaningful models
- Design scalable algorithms for social data processing
- Utilize networks and graphical models
- Obtain qualitative insights from dynamic network models
Who Should Attend
Statisticians, data scientists, marketing officers (CMOs), data analytic experts, researchers in social sciences, faculty in social sciences and marketing, and political bloggers.
Day 1. Social data, networks, and graphs
Overview of social data
- What's available: private vs. public, e.g. Facebook, Twitter, Yelp, blogs, etc.
- What we can expect to get out of this data, e.g. examples from popular media such as ‘Tipping Point’, successful marketing campaigns, political campaigns, and societal revolutions.
- Why they are networked and why the existence of a network plays an important role.
Models of social networks
- Connection level models: Stochastic/random graph models, static vs. dynamic, branching processes
- Behavioral models over networks: SIR model, Voter model
- Operational/functional network model: Probabilistic graphical models, Graph of Users and Items for recommendation
- Graph properties: Degree, diameter, between-ness, conductance, PageRank, clustering, etc.
Day 2. From data to decisions via prediction
We will focus on four cases today to illustrate how we use data plus domain knowledge to infer the model, which allows us to make appropriate predictions, leading us to eventual decisions.
- Searching the web - PageRank centrality
- Identifying influential agents or information sources - Rumor centrality / Viral marketing and centralities
- Recommendation - Building the user-item graph / Learning the hidden preferences
- Discovering communities - Learning the graph / Clustering
Discussion / Challenges
The day will wrap up with a discussion of the challenges related to computation, 'data cleaning,' and qualitative information extraction.
Course schedule and registration times
Registration is on Monday morning from 8:15 - 8:45 am.
Class runs 9:00 am - 5:00 pm on both days.
Note that a laptop is required for this course, preferably with Python installed.
About The Lecturers
Devavrat Shah is currently a Jamieson career development associate professor with the department of Electrical Engineering and Computer Science, MIT. He is a member of the Laboratory for Information and Decision Systems (LIDS) and Operations Research Center (ORC). His current research interests include social networks and statistical inference.
His work has been recognized with awards for papers in machine learning, networking, and operations research; the 2008 ACM Sigmetrics Rising Star award and 2010 INFORMS Erlang Prize; and recent press releases in The New York Times (on recommendation systems) and Forbes (on predicting trends in Twitter).
Professor Dahleh is a professor in and the Associate Department Head of the Department of Electrical Engineering and Computer Science at MIT. Previously, he was the acting director of the Laboratory for Information and Decision Systems.
Dr. Dahleh is interested in problems at the interface of robust decisions, filtering, information theory, networks, and computation. In addition to methodology development, he is interested in the application of distributed control in the future electric grid and the future transportation system with particular emphasis in the management of systemic risk. He is also interested in various problems in social networks including information propagation over complex networks, global games and application to revolution models, and various connections to health care.
This course takes place on the MIT campus in Cambridge, Massachusetts. We can also offer this course for groups of employees at your location. Please contact the Short Programs office for further details.
Links & Resources
- Devavrat Shah on Twitter trends Shah discusses how he uses his machine-learning algorithm to predict trends hours in advance of recognition by Twitter.
- Predicting what topics will trend on Twitter A new algorithm predicts which Twitter topics will trend hours in advance and offers a new technique for analyzing data that fluctuate over time.