● Train and score 2-way bond pricing models in real-time
● Predict errors between price and rule based predictions
● Optimize rules engine with greater accuracy
● Improve accuracy of reference price predictions on days with high variability in the market
● Predict OAS (Option Adjusted Spread) for Corporate Bonds
● Improved performance, scoring 24,000 corporate bonds globally for a single customer
● Predicted pricing accuracy every 15 seconds, removing features that were not predictive to price
● Reduced errors by $0.66 per bond
● The AI model will predict the 90 days forward OAS for the corporate bonds.
● Feature includes issuer fundamentals, current credit ratings, bond features and other fixed income analytics variables.
● The solution can be delivered through APIs to the research team.
● Fulfill client liquidity needs with more accurate and timely bond pricing data
● Predicting OAS (Option Adjusted Spread) is a good indicator that can be used to assess the credit quality of the bond and an important input to determine the bond ratings (high yield to investment grade or vice a versa)
H2O's AI and Data Approaches
● Classification Models
● H2O-3 gradient boosting machine to improve performance
● Automatically create interactions between features
● Perform feature selection to remove features that are not predictive to price