Using Deep Learning Techniques to Predict 10-Year US Treasury Yield

Abstract

The yield to maturity of United States Treasury securities is a decisive indicator of the economic cycle in the United States, and it is also one of the most critical interest rate references for capital markets worldwide. This study investigates the effectiveness of applying deep learning methods in financial prediction. Specifically, a deep learning model is trained by using the yields of various United States Treasury securities of different maturities to predict the 10-year yield.We collect time series data from the daily yields of United States Treasury securities from January 1990 to November 2018, which are subsequently preprocessed for the establishment of a long short-term memory model. By using this model, we predict the 10-year yield with a resulting mean squared error as low as 0.0063 for the test data sets.