Adaptive Combination of Diverse Models for Government Bond Yield Curve Forecasting

Abstract

Forecasting the government bond yield curve remains a relevant task in financial analytics. Comparative studies indicate the absence of a single model that consistently outperforms alternatives across all maturities and market conditions. Traditional econometric approaches demonstrate greater robustness under volatile conditions, while machine learning models provide higher accuracy in stable environments. This trade-off between accuracy and robustness justifies the use of combined approaches.