All or most of the work includes a mix of the following technologies: R, R libraries, R Shiny, APIs, MySQL, Neo4j, MSSQL, Linux.
Clinical decision support system, a fully functional, end-to-end proof of concept using text mining, machine learning, graphs and graph database (neo4j), bioinformatics and medical ontologies. The idea is to bring artificial intelligence to mimic how a brain learns and – equally important – forgets.
- Multiple physicians found the CDSS very useful, very flexible, incredible fast and fun to use
- More details in Clinical Decision Support System.
Argument Graphs from scientific journals in PDF format, finding references to supporting arguments, bioinformatics, images and data in intra and inter documents, and visualize reference trees in an interactive argument graph
- Multiple APIs used including to PubMed, CrossRef and Orcid
- Data streams merged into a graph, stored in a graph database (neo4j)
Oncologist Networks using graph community detection algorithms to detect oncologist communities using graph database (neo4j) and R
- Each oncologist network is ranked based on importance, then compared to find the strongest network of oncologists
- The idea is that today, we are no longer looking for the best physician, instead, we are looking for the best physician network
- Work also included prediction and identification to find the up and coming future Oncologists super stars
- Physician network detection is important for patients, payers and the pharmaceutical industry
Identifying key opinion leaders in Oncology; specific therapy areas; mashup of data streams from a variety of sources
- PubMed, ClinicalTrials.gov and private databases
- Visualization using non-interactive and interactive graphs (java script)
End-to-end Twitter analysis from reading Twitter (API) to visualization
- Including n-grams, association rules, geolocation, keyword trending, interactive visualization and graph theory
- Graph database (neo4j)
Cryptocurrency analysis, using natural language processing (NLP) of Twitter and Reddit data in cryptocurrency
- Research to anticipate cryptocurrency price and volume fluctuations based on social media chatter
Electronic nordic skipole handle analytics
- Left and right ski handles send sensor data to a phone app (50 times / sec). The data is uploaded to the cloud for analysis
- Built analytics using R and Monet DB (for very fast transfer of large amount of data)
- Using R Shiny to build a front end