BEGIN:VCALENDAR
PRODID:-//hacksw/handcal//NONSGML v1.0//EN
METHOD:PUBLISH
BEGIN:VEVENT
DESCRIPTION:Click for Latest Location Information: http://smartdata2015.dataversity.net/sessionPop.cfm?confid=91&proposalid=8174\nDuring this conference we are introducing the “Semantic Data Lake” for Health Care which at one hospital contains 10 years of medical records for close to 3 million patients. For each patient there are time stamped diagnoses, procedures, prescriptions, and all relevant demographic variables. The data is grounded in a unified taxonomy and ontology that ties billing codes to all of the important medical terminology systems (i.e. - MESH, SnoMed, OMOP, RxNorm, UMLS, loinc, etc).\nThe Semantic Data Lake stores the information about patients in Hadoop and is indexed with a Semantic Graph database. Additionally, we provide data mining and machine learning capabilities via R or SPARK. This architecture allows data scientists to perform very advanced queries and feature selection against the data without having to write any map/reduce or procedural SPARK code and then continue their advanced analytics with the tools they prefer.\nIn the lightning talk we will demo the Semantic Data Lake’s powerful combination of predictive analytics with a Semantic Graph database. Using a matrix of likelihood measures between any two diagnoses based on temporal ordering and then added that matrix back in the Semantic Graph database, we create what we have termed a “Probability Graph”. This approach allows data scientists in health care or the insurance industry a radical new way to navigate through a space of diseases, drugs and procedures. \nFor example: a health care professional can predict given your diagnoses in the past what your most likely next diseases are going to be.  Insurance professionals can find unexpected new diagnoses or procedures that might point to provider fraud or member fraud.\nThe same Semantic Data Lake and the Probably Graph approach has capabilities in the intelligence domains to predict the likelihood of a person’s next behavior.   We also explore the opportunities to use the same approach in on-line learning where one wants to predict future performance and knowledge gaps based on actual measurements of student behavior.
DTSTART:20150818T160000
SUMMARY:What Disease is in Your Future - Exploring the Probability Graph
DTEND:20150818T162959
LOCATION: See Description
END:VEVENT
END:VCALENDAR