Analysis of Interest Rate Curves Clustering Using Self-Organising Maps

The paper presents the analysis of temporal evolution of interest rate curves (IRC). IRC are considered as objects embedded in a high-dimensional space composed of 13 different maturities. The objective of the analysis was to apply a nonlinear nonparametric tool (Self-Organising Maps) to study the clustering of IRC in three different representations: the original curves, the increments and 3-parametric Nelson-Siegel model. An important finding of this study reveals the temporal clustering of IRC behaviour which is related to the market evolution. Other results include the relative analysis of CHF-EUR evolution and the clustering found in the evolution of factors used by Nelson-Siegel model. The analysis of the consistency of these factors to represent the typical IRC behaviour requires further work. Current results are useful for the development of interest rates forecasting models and financial risk management.