DATA SCIENCE TEA

09 Nov
Monday, 11/09/2015 4:00pm to 5:30pm
Computer Science Building, Room 150 & 151
Special Event

Tea, refreshments, presentations and conversations about topics in data science. 

This week's presenters include: 

Stephen A Lauer
(PhD Student in Biostatistics advised by Prof. Nicholas Reich)
Title: Real-time forecasting dengue hemorrhagic fever in Thailand

Abstract: Dengue is a mosquito-borne infectious disease that places an immense public health and economic burden upon Thailand. In collaboration with the Thai Ministry of Public Health and Bureau of Epidemiology, our group has developed a statistical model for infectious disease surveillance that uses data from across Thailand to give early warning of developing dengue epidemics in real time. For each province, the forecast is based on seasonal dynamics of dengue in the focal province and observed case counts at recent time-points from the focal province and neighbors demonstrated to be relevant through model selection using historical data. I will present the challenges we faced and the results of this forecasting exercise in predicting the 2014 dengue season in each Thai province.

Joanna Asia Biega
(Max Planck Institute for Informatics, PhD student advised by Gerhard Weikum)
Title: IR-Centric Privacy Risk Assessment in Online Communities
Abstract: Privacy of Internet users is at stake because they expose personal information by posts in online communities, queries, and other textual contents. An adversary applying Information Retrieval (IR) systems to user data and comparing the data with the rest of the community may infer sensitive properties and utilize them against the user (for instance, by issuing irritating ads or adjusting the pricing of goods or services). We propose a privacy risk assessment model representing adversaries and their background knowledge using some of the IR techniques, and evaluate its performance against crowdsourced adversarial judgements.

Evan Ray
(Post-doctoral fellow working with Prof. Nicholas Reich)
Title: Predicting infectious disease incidence with kernel density estimation

Abstract:  I discuss the use of kernel density estimation (KDE) to obtain predictive distributions for the trajectory of infectious disease incidence over a sequence of multiple prediction horizons.  I briefly outline some limitations of a direct application of common implementations of KDE to this prediction task and some partial remedies for these problems.

In case of questions, contact: Nicholas Monath (nmonath@cs.umass.edu).