Journal of Computational Finance

Risk.net

Yield curve fitting with artificial intelligence: a comparison of standard fitting methods with artificial intelligence algorithms

Achim Posthaus

  • In this paper, the author expands standard yield curve fitting techniques to artificial intelligence methods.
  • AI methods show a comparable goodness:
    - of the yield curve fit to Overnight Index Swap data (RMSE test)
    - interpolation quality (Leave-one-out test)
  • The author broadens AI methods to multi-dimensional fits incorporating time series information.
  • Extension of fitting methods to AI algorithms may serve in practice as an:
    - additional way to calibrate yield curves
    - independent cross check and validation of results of standard methods

The yield curve is a fundamental input parameter of valuation theories in capital markets. Information about yields can be observed in a discrete form, either directly through traded yield instruments (eg, interest rate swaps) or indirectly through the prices of bonds (eg, government bonds). Capital markets usually create benchmark yield curves for specific and very liquid market instruments, or for issuers where many different quotes of individual yield information for specific maturities are observable. The standard methods to construct a continuous yield curve from discrete observable yield data quotes are the fit of a mathematical model function, interpolation or regression algorithms. This paper expands these standard methods to include artificial intelligence algorithms, which have the advantage of avoiding any assumptions with regard to the mathematical model functions of the yield curve, and which can conceptually adapt easily to any market changes. Nowadays, the most widely used risk-free yield curve in capital markets is the overnight index swap (OIS) curve, which is derived from observable OISs and is used in this paper as the benchmark curve to derive and compare different yield curve fits.

To continue reading...

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: