Libor Concept and Libor’s Estimation With Multi-Layer Perceptron

Authors : Ahmet Yüzbaşıoğulları

Pages : 34-53

Doi:10.58724/assam.1702558

View : 50 | Download : 169

Publication Date : 2025-11-12

Article Type : Research Paper

Abstract :Economic developments since the 1970s and excessive volatility in interest rates have necessitated new solutions for long-term borrowing. In this context, the London Interbank Offered Rate (LIBOR), a comparison tool for variable interest rates, has been an important solution. The use of LIBOR, calculated as a reference interest rate by the BBA in 1986, increased until the 2008 Global Financial Crisis but was subsequently questioned due to post-crisis manipulation allegations and replaced in the 2020s by SOFR in the US, SONIA in the UK, and ESTR in the EU. The estimation of the future value of LIBOR, which determined interbank borrowing costs for many years, is of great importance for companies and countries. Interest rate modeling began with Louis Bachelier\\\'s 1900 work on Arithmetic Brownian Motion (ABM). Although the LIBOR Market Model (LMM) is widely used today, its application is limited due to its complexity, calibration problems, simulation requirements, and the discontinuation of LIBOR. In this study, traditional econometric methods were insufficient as interest rate movements were not linear and parametric. Therefore, the Multi-Layer Perceptron (MLP) method, an artificial intelligence application, was used. Due to reasons such as the lack of sufficient SOFR data, one-month LIBOR was taken as the dependent variable and the FED policy rate as the independent variable. Both cointegration and MLP analyses reveal that LIBOR moves one-to-one with the FED policy rate. The correct estimation of the FED Policy Rate will enable an accurate estimation of the future LIBOR rate.
Keywords : Londra Bankalararası Teklif Faizi (LIBOR), LIBOR Piyasa Modeli (LMM), Çok Katmanlı Algılayıcı (MLP)

ORIGINAL ARTICLE URL