OTCStreaming ambition is to become the market standard source for Credit Derivatives valuations.
We are receiving data from different sources. On one hand from the Standard Repository Data (SDR) where, as instructed by regulators, every market participant has to upload his trades on cleared instruments (like credit indices) and from multiple not formally formatted messages sent by traders with indication of interests.
The challenge was to be able to extract in real time the data out of this multiple non standardized messages. Then to classify and enrich the data to be able to publish it and calculate closing prices and theoretical valuations.
We believe that machine learning strategies with a high F-score is the most flexible solution to extract data and information from specialized texts. ScriptMiner extracts quantitative data and qualitative information from technical email content. We have applied our technology to the credit market emails.
We plan to deploy ScriptMiner across many industries outside of its current focus on the financial market: we are targeting professional emails with specific data content.
Introducing the machine learning technology in expert systems reduces the maintenance costs and improves systems' accuracy.
CD Suite is a front to back solution for trading and managing fixed income products (credit derivatives and corporate bonds) on a real time basis. The software architecture is fully compliant with all new regulatory requirements.
CD Suite first release in 2013 was mostly an expert system but already demonstrated machine learning capabilities. Risks were computed according to our cluster-based model of time series. Since 2013, we are investing in making the system more clever regarding all its interactions on data, reporting and processes. Opposite to existing solutions, our architecture makes it very flexible, one can add a new instrument class and the cluster analysis will capture it to run risk analysis with minimum efforts.
The copula framework provides a very stable cluster representation of financial time series.
DataGrapple lets you explore supervised and unsupervised clusters of time series. We own a proprietary database of credit derivatives prices since 2006. The database includes over 700 reference entities. The project objective was to characterize the joint behaviour of this large and complex database. Using recent research works, we have designed a robust model using the copula representation of the joint behaviour of CDS. The clusters are very stable over time compared to a standard cluster analysis. We have introduced the Grapple, a user friendly innovation in data representation. A Grapple is an expert best selection of a cluster, a graphic representation and a time series.