Marina Sokolova
University of Ottawa and
Institute for Big Data Analytics, Dalhousie University
ABSTRACT:
Data holders in healthcare, finance and other sectors increasingly collect
person-specific information. These data sets are routinely used by Machine
Learning to predict patient's diagnosis, credit scores, etc.
In this talk we consider a reverse problem of classification of
demographic characteristics from given data sets. In other words, we
look at how Machine Learning can analyze personal data from the
perspective of Social Mining. We use Multi-Label Classification to
classify/predict demographic characteristics of patients diagnosed with
Diabetes. We report the results of age, race and gender
classification obtained on data collected at US hospitals. The work is
done with Dung Tran, Naveen Kumar, Aravind Doss.