The present study deals with the empirical analysis of patterns in the evolution of interest rate curves (IRC). The main topic is to consider IRC as objects (curves) embedded into high-dimensional space and to study similarities and differences between them. This is a typical problem of clustering and classification in machine learning. In fact, theses data – IRC, can be considered as functional data set. Machine learning algorithm, namely Self-Organising Map SOM (Kohonen map) [1], was used to study the evolution of interest rates and to reveal the potential patterns and clusters in IRC. Case study is based on Swiss franc (CHF) data on daily interest rates. For the analysis both raw data (curves composed of 13 non-regularly distributed maturities from 1 week to 10 years) and data completed by interest rates mapping in a feature space of date-maturity were studied [2]. In the latter case curves are composed of 120 regularly (by month) distributed maturities. Feasibility study and preliminary results on IRC patterns analysis first time were presented in [3].
[1]
T. D. Matteoa,et al.
An interest rates cluster analysis
,
2004
.
[2]
F. Diebold,et al.
Forecasting the Term Structure of Government Bond Yields
,
2002
.
[3]
V. Timonin,et al.
Classification of Interest Rate Curves Using Self-Organising Maps
,
2007,
0709.4401.
[4]
Tomaso Aste,et al.
Interest rates hierarchical structure
,
2005
.
[5]
Alexei Pozdnoukhov,et al.
Interest rates mapping
,
2007,
0709.4361.