Asset Returns Interval Forecasting: A Singular Spectral Analysis-Based Varying-Coefficient Autoregressive Conditional Interval Model  
Author Xingyu Dai

 
Co-Author(s) Roy Cerqueti; Dongna Zhang; Qunwei Wang

 
Abstract As a response to the complex variations of asset prices, the traditional point-valued forecasting exercise may not satisfy investors' risk management needs. Using a random set and singular spectral analysis theory, this paper proposes a novel varyingcoefficient autoregressive conditional interval (VC-ACI) model and a modified interval-valued singular value decomposition (ISVD) method to construct a mISVD-VC-ACI decomposition-ensemble model. The proposed model can be used to forecast interval-valued returns. The empirical study selects WTI and Brent oil futures weekly data from January 2, 2000, to March 22, 2023, and chooses 20 different benchmark models. This paper finds that firstly, the mISVD-VC-ACI model outperforms all benchmark models in a one-step ahead forecasting scenario. The model set with the mISVD decomposition methods and the VC-ACI forecasting model also outperform the benchmark model set. Secondly, when the number of forecasting steps is increased, the prediction performance superiority of the ISVD-VC-ACI model becomes more distinct. Thirdly, when the forecasting rolling window length is increased, the mISVD-VC-ACI model demonstrates robust interval-valued prediction performance. This study also provides a novel paradigm for conducting crude oil futures interval-valued returns pattern analyses.

 
Keywords
   
    Article #:  DSBFI25-67
 
Proceedings of 3rd ISSAT International Conference on Data Science in Business, Finance and Industry
January 6-8, 2025 - Da Nang, Vietnam