Targeted Learning in Healthcare Research

Susan Gruber, Jan 15, 2016
Commentary by Stephen Downes

This article is focused on the big data in electronic health records and claims data sets being put to secondary use in studying questions of drug safety and efficacy, but it applies equally to learning analytics. The idea of 'targeted learning' is to draw from machine learning algorithms (which would be engaged in pattern matching and cluster detection, for example) which do not use parameters ("these avoid model misspecification bias by making no distributional assumptions") but to supplement it with targeted parameters to answer specific questions from the data. The article is fairle accessible and discusses methodologies and applications in a relatively short 8 page PDF. View more from this special issue of Big data on healthcare and data. See also Mining the Quantified Self, which is really an excellent overview of the needs and challenges facing personal data analytics. (Note: I have full access to these in my office; if they're subscription-based, please accept my apologies).


Views: 1 today, 278 total (since January 1, 2017).[Direct Link]