Despite the large number of bonds in the universe, most bonds trade very infrequently (7 trades per CUSIP on average, annually). This reality can make pricing on a bank’s bond inventory difficult as there is little recent data in which to support a decision. If one could identify similar bonds to the one that is being priced, they could be used as a reference to better determine market value. However, translating how a human determines “similarity” into math that a model can use to make predictions is a significant challenge
Model Deployed in Business
ModelOp operationalized the life cycle of a Similarity Model that accurately predicts pricing information for municipal securities in a way that is transparent and interpretable by traders. The pricing information also includes a confidence score indicating level of certainty for the prediction. The data acquisition, execution, and data publication of this model is automated; creating a seamless experience for the trader through a web dashboard (Dash).
ModelOp helped this customer deploy its first Docker environment in which to standardize the model deployment platform. The Docker environment allows for flexibility and control as new technologies are introduced into the solution.
Key Data Sources
Market Trade Data
For Businesses, Not Scientists
Our focus on model operations has established a mature, repeatable process in which to deploy new analytics products to business users. A process that took months can now be accomplished in weeks. In addition, the pricing model has solved a long running business problem of identifying similar securities with available sparse data set. This model provides interpretable output for pricing guidance, giving traders a confident recommendation in which to make decisions.
Explore the Impact of ModelOps
We have the experience and solution to help you harness the power of AI/ML at scale that will boost your core processes. Let’s create a unique solution together.