Machine Learning applied to Municipal Scale Interpolation and Extrapolation

A team of finance and technology professionals at Lumesis have harnessed artificial intelligence tools to begin solving practical problems that arise when developing deal-specific yield curves (or scales) during the new municipal bond issuance pricing process. Specifically, Lumesis has developed a methodology to predict the shape of a deal-specific municipal yield curve based on the unique structural and credit characteristics of the deal being priced and to use that predicted shape to fill-in portions of a yield curve that cannot be built from observable market transactions. The artificial intelligence-based solution that has been developed is described below, along with the results of empirical and statistical backtesting. The results indicate a significant advance in the ability to generate complete deal-specific municipal yield curves over traditional interpolation and extrapolation methods.