Banca de QUALIFICAÇÃO: GUILHERME COSTA CHADUD MOREIRA

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : GUILHERME COSTA CHADUD MOREIRA
DATE: 02/02/2026
TIME: 17:00
LOCAL: A DEFESA SERÁ REALIZADA POR VIDEOCONFERÊNCIA
TITLE:

Identifying early warning signals in the Brazilian nominal yield curve using
neural networks

 

KEY WORDS:

Yield Curve; Neural Networks; Warning System; LSTM.

PAGES: 51
BIG AREA: Ciências Sociais Aplicadas
AREA: Economia
SUMMARY:

This project aims to identify early warning signals in the Brazilian nominal
yield curve over the period from 2015 to 2025, using the Long Short Term Memory (LSTM)
machine learning technique. The initial premise of the study is to examine the presence of
anomalous behavior in market interest rates in order to detect, within the temporal structure of
the data, points that deviate from the recent pattern of the maturities under analysis. In this
context, an Early Warning System (EWS) is developed using LSTM neural networks, applied
to the dynamics of the nominal yield curve across multiple maturities, employing a rolling
window approach. Consequently, the project contributes to the debate on the ability to identify
signals of change in the yield curve, providing potential early indications to market participants
in an environment characterized by significant predictive challenges of different natures.

 

COMMITTEE MEMBERS:
Interno - 1642911 - DANIEL OLIVEIRA CAJUEIRO
Externo à Instituição - Douglas Sad Silveira - UFJF