Long-term Forecasting of Risk Indicators for Chinese Financial Derivatives Market Based on Seasonal-trend Decomposition and Sub-components Modeling

Financial market monitoring has been becoming a research hotspot in the past few years. A solution to the problems of immediate risk predicting and future health status estimating is meaningful to the health management of the market. Researchers have achieved impressive results in digging for valuable information from short-term financial time series, while long-term series forecasting still keeps a challenging task. This challenge originates from the complexity and uncertainty of the financial market, which involves investors’ unpredictable behaviors, the fluctuant stock market, and even off-market factors. Traditional models usually fail to grasp long-term changes in the financial market. This paper focuses on one of the most rapidly-changing financial markets, namely the derivatives market. The target is to predict long-term changes in two widely-used risk indicators including trading volume and open interest. The proposed approach consists of two steps that are referred to as seasonal-trend decomposition and sub-component modeling. The historical series is first decomposed into seasonal and trend components. Then a FEDformer model and a frequency-enhanced LSTM model are used to predict long-term changes in the sub-components, respectively. This algorithm has been applied to the time series from the derivatives market in China. Results show that the proposed framework can predict the long-term changes in risk indicators at higher accuracy compared with Transformer-based methods.