Sunday 3:50 PM–4:35 PM in Track 3

Visualizing Machine Learning of Units of Measure using PyViz

Jacob Barhak

Audience level:
Intermediate

Description

This talk will discuss how PyViz tools were used to visualize Machine Learning employed to learn clinical units of measure from ClinicalTrials.Gov and multiple standards

Abstract

By: Jacob Barhak & Joshua Schertz Clinical trials are now reported in ClinicalTrials.Gov according to U.S. law that requires registration of clinical trial data. This NIH/NLM governed registry provides important modeling data information by accumulating over 300,000 clinical trials. However, despite the great effort by the government to centralize the data, the reporting entities do not follow a standard when entering data. Therefore, the information is machine readable, yet not machine comprehensible. As of 12 Apr 2019, all 35,926 trials with results had 24,548 different units. The authors created a solution based on machine learning to address this standardization problem. The solution includes:
* Unsupervised machine learning * A web site to provide mapping ClinicalUnitMaping.Com * Representation of multiple unit standards: CDISC, NIST / RTMMS / IEEE, Unit Ontology / Bio Portal, UCUM * Supervised machine learning of units.

This publication will focus on visualization of unsupervised and supervised learning employed in this project using the PyViz library while explaining how other python libraries were used in this project.

Subscribe to Receive PyData Updates

Subscribe