"Recursive Multistep Forecasting of U.S. Treasury Yield Curves Using Ma" by Aman Singh

Author

Aman Singh

Date of Award

9-9-2024

Document Type

Thesis

Publisher

Santa Clara : Santa Clara University, 2024

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Advisor

Tokunbo Ogunfunmi

Abstract

While machine learning has revolutionalized many fields such as natural language processing (NLP) and computer vision, its impact on time series forecasting is still widely disputed. This thesis focuses on comparing forecasting performance between econometrics/time series analysis, classical machine learning, and deep learning methods. More specifically, it is recursive multistep forecasting of the U.S. Treasury daily yield curve. The U.S Treasury daily yield curve is important because it is considered a leading indicator of recessions. While the majority of yield curve forecasting research focuses on monthly data, we are focusing on business daily data. We have also included a wide variety of methods that have not been tested on yield curve forecasting, especially deep learning algorithms. The algorithms include the Autoregressive Integrated Moving Average (ARIMA) model and its extensions, naive benchmarks, ensemble methods, Recurrent Neural Networks (RNNs), and multiple transformers built for forecasting. The thesis finds that ARIMA and naive benchmarks outperform all other models overall, except in one time period. Of the machine learning methods, TimeGPT and RNNs perform the best. Furthermore, the thesis explores whether stationary or nonstationary outputs are more appropriate as input to deep learning models.

Share

COinS