How Do You Apply Machine Learning to Deliver New Functionality For a Complex Application Given Limited Time?

GOTO Amsterdam 2017

How do you apply Machine Learning to deliver new functionality for a complex application given limited time? We did just that for GeriMedica, a company that builds medical record platforms for medical professionals. GeriMedica wanted to gain more insight into the treatment process through recommendations at crucial decision points. We succeeded through Machine Learning, and found causality in semi-structured data in over 221 million events of patient interactions.

The biggest challenge we faced was delivering Machine Learning features on a tight deadline. In this talk, we explain how we applied Machine Learning in an existing application in an agile way. We had to complete relevant business questions, important data sources, and integration within the application landscape in a matter of weeks. We used Word2Vec and Doc2Vec to pre-process the text and get initial insight, after which we demonstrated that we could use Keras to learn to recognize typical cases.