Cognitive Probability Graphs for Knowledge Management


​Franz Inc. is collaborating with the Montefiore Health System, Intel, Cloudera, and Cisco to deploy a cognitive computing platform for healthcare. The platform is used in the medical domain for personalized medicine, translational research, predictive modeling, real time decision support and most importantly: computing the true cost of care. The first version of the platform holds ten years of data for 2.7 million patients and is currently in production for several use cases.

Our semantic healthcare platform ingests a wide variety of patient data, ranging from EMR data to real time data from sensors in the ICU. All data is transformed into a unified clinical event model and all the terms in the model are linked to a unified terminology system that consists of more than 182 vocabularies and taxonomies in healthcare and the life sciences. On top of that we integrate external knowledge bases such as clinical trial databases, drugs and disease databases, GO, Pubmed, etc to allow for reasoning and advanced querying capabilities.
Our cognitive platform is implemented using a distributed semantic graph database that is queried via a parallel SPARQL implementation.  The output of queries are pipelined into data science packages such as R, SPARK-ML, and  H2O and the analytic results are fed back as a graph into the platform.  This way the output of analytics becomes a queryable part of our infrastructure and it also becomes straightforward to visualize the results of your analytics. We call this approach the cognitive probability graph and in this presentation we will demo the various aspects of both the platform and the use of probability graphs.
Outside of healthcare we also actively explore a number of applications for the cognitive probability graph where the inclusion of probability, similarity and uncertainty as a graph in your data deliver novel solutions. In the presentation we will give examples and demos from Ecommerce, police intelligence and logistics.