Big Data Analytics for Economic Development

Abstract:
The Web is a collection of vast and diverse datasets, like news
articles, user posted articles, (e.g. blogs), social media
(e.g. twitter), images, in addition to different governmental and
non-governmental organization post structured data on their
websites. My research is focused on building a generic big data
analytics engine that can analyze parallel streams of diverse and
noisy datasets from the Web to infer and forecast macroeconomic and
societal indicators leading to a more informed decision-making and
better management of resources and events.  Presently, the main focus
of my work is to extract events and infer spatio-temporal relationship
between events using online news articles and Web media. Such analysis
can help to understand hitherto unseen or unknown correlated
socio-economic phenomenon. For example, is there a relationship
between bad monsoon and rise in food prices?  Our current system can
produce temporal event relationship maps for news articles and we also
have early evidence of temporal causal relationships between events
(e.g., monsoon levels influence dengue intensity).  From temporal
causal dependencies, we aim to learn important events to track for a
given macroeconomic or societal phenomena and use these events to
learn a Bayesian network structure for predicting a location specific
index for any given phenomena.

Bio:
Sunandan Chakraborty is a PhD candidate with the Computer Science
department of New York University. His research interests include data
mining, information retrieval, web search and computing for development.
He is presently doing an internship in the Microsoft Research Cambridge.